500 MB Nyheder Direkte fra Internet 2/500 MB nyheder direkte fra internet CD 2.iso/start/data/text/faq-1005.txt

Text File | 1995-05-08 | 165KB | 3,490 lines

Archive-name: neural-net-faq
Last-modified: 1995/04/20
URL: http://wwwipd.ira.uka.de/~prechelt/FAQ/neural-net-faq.html
Maintainer: [e-mail] (Lutz Prechelt)


  ------------------------------------------------------------------------
        Additions, corrections, or improvements are always welcome.
        Anybody who is willing to contribute any information,
        please email me; if it is relevant, I will incorporate it.

        The monthly posting departs at the 28th of every month.
  ------------------------------------------------------------------------


This is a monthly posting to the Usenet newsgroup comp.ai.neural-nets
(and comp.answers, where it should be findable at ANY time). Its
purpose is to provide basic information for individuals who are new to the
field of neural networks or are just beginning to read this group. It shall
help to avoid lengthy discussion of questions that usually arise for
beginners of one or the other kind.

   SO, PLEASE, SEARCH THIS POSTING FIRST IF YOU HAVE A QUESTION
                           and
   DON'T POST ANSWERS TO FAQs: POINT THE ASKER TO THIS POSTING

This posting is archived in the periodic posting archive on host
rtfm.mit.edu (and on some other hosts as well). Look in the anonymous
ftp directory "/pub/usenet/news.answers", the filename is as given in the
'Archive-name:' header above. If you do not have anonymous ftp access,
you can access the archives by mail server as well. Send an E-mail
message to [e-mail] with "help" and "index" in the
body on separate lines for more information.

For those of you who read this posting anywhere other than in
comp.ai.neural-nets: To read comp.ai.neural-nets (or post articles to it)
you need Usenet News access. Try the commands, 'xrn', 'rn', 'nn', or 'trn'
on your Unix machine, 'news' on your VMS machine, or ask a local
guru.

This monthly posting is also available as a hypertext document in WWW
(World Wide Web) under the URL
"http://wwwipd.ira.uka.de/~prechelt/FAQ/neural-net-faq.html"

The monthly posting is not meant to discuss any topic exhaustively.

Disclaimer:
   This posting is provided 'as is'.
   No warranty whatsoever is expressed or implied,
   in particular, no warranty that the information contained herein
   is correct or useful in any way, although both is intended.

To find the answer of question number 'x', search for the string
"x. A:" (so the answer to question 12 is at   12. A:  )


And now, in the end, we begin:

========== Questions ==========
********************************

 1. What is this newsgroup for? How shall it be used?
 2. What is a neural network (NN)?
 3. What can you do with a Neural Network and what not?
 4. Who is concerned with Neural Networks?

 5. What does 'backprop' mean? What is 'overfitting'?
 6. Why use a bias input? Why activation functions?
 7. How many hidden units should I use?
 8. How many learning methods for NNs exist? Which?
 9. What about Genetic Algorithms?
 10. What about Fuzzy Logic?
 11. How are NNs related to statistical methods?

 12. Good introductory literature about Neural Networks?
 13. Any journals and magazines about Neural Networks?
 14. The most important conferences concerned with Neural
   Networks?
 15. Neural Network Associations?
 16. Other sources of information about NNs?

 17. Freely available software packages for NN simulation?
 18. Commercial software packages for NN simulation?
 19. Neural Network hardware?

 20. Databases for experimentation with NNs?

========== Answers ==========
******************************

 1. A: What is this newsgroup for? How shall it be
 =================================================
   used?
   =====

   The newsgroup comp.ai.neural-nets is intended as a forum for
   people who want to use or explore the capabilities of Artificial
   Neural Networks or Neural-Network-like structures.

   There should be the following types of articles in this newsgroup:

    1. Requests
    +++++++++++

      Requests are articles of the form "I am looking for
      X" where X is something public like a book, an article, a
      piece of software. The most important about such a request
      is to be as specific as possible!

      If multiple different answers can be expected, the person
      making the request should prepare to make a summary of
      the answers he/she got and announce to do so with a
      phrase like "Please reply by email, I'll
      summarize to the group" at the end of the posting.

      The Subject line of the posting should then be something
      like "Request: X"

    2. Questions
    ++++++++++++

      As opposed to requests, questions ask for a larger piece of
      information or a more or less detailed explanation of
      something. To avoid lots of redundant traffic it is important
      that the poster provides with the question all information
      s/he already has about the subject asked and state the
      actual question as precise and narrow as possible. The
      poster should prepare to make a summary of the answers
      s/he got and announce to do so with a phrase like
      "Please reply by email, I'll summarize to
      the group" at the end of the posting.

      The Subject line of the posting should be something like
      "Question: this-and-that" or have the form of a
      question (i.e., end with a question mark)

    3. Answers
    ++++++++++

      These are reactions to questions or requests. As a rule of
      thumb articles of type "answer" should be rare. Ideally, in
      most cases either the answer is too specific to be of general
      interest (and should thus be e-mailed to the poster) or a
      summary was announced with the question or request (and
      answers should thus be e-mailed to the poster).

      The subject lines of answers are automatically adjusted by
      the news software. Note that sometimes longer threads of
      discussion evolve from an answer to a question or request.
      In this case posters should change the subject line suitably
      as soon as the topic goes too far away from the one
      announced in the original subject line. You can still carry
      along the old subject in parentheses in the form
      "Subject: new subject (was: old subject)"

    4. Summaries
    ++++++++++++

      In all cases of requests or questions the answers for which
      can be assumed to be of some general interest, the poster of
      the request or question shall summarize the answers he/she
      received. Such a summary should be announced in the
      original posting of the question or request with a phrase
      like "Please answer by email, I'll
      summarize"

      In such a case, people who answer to a question should
      NOT post their answer to the newsgroup but instead mail
      them to the poster of the question who collects and reviews
      them. After about 5 to 20 days after the original posting, its
      poster should make the summary of answers and post it to
      the newsgroup.

      Some care should be invested into a summary:
       o simple concatenation of all the answers is not
         enough: instead, redundancies, irrelevancies,
         verbosities, and errors should be filtered out (as good
         as possible)
       o the answers should be separated clearly
       o the contributors of the individual answers should be
         identifiable (unless they requested to remain
         anonymous [yes, that happens])
       o the summary should start with the "quintessence" of
         the answers, as seen by the original poster
       o A summary should, when posted, clearly be
         indicated to be one by giving it a Subject line
         starting with "SUMMARY:"
      Note that a good summary is pure gold for the rest of the
      newsgroup community, so summary work will be most
      appreciated by all of us. Good summaries are more valuable
      than any moderator ! :-)

    5. Announcements
    ++++++++++++++++

      Some articles never need any public reaction. These are
      called announcements (for instance for a workshop,
      conference or the availability of some technical report or
      software system).

      Announcements should be clearly indicated to be such by
      giving them a subject line of the form "Announcement:
      this-and-that"

    6. Reports
    ++++++++++

      Sometimes people spontaneously want to report something
      to the newsgroup. This might be special experiences with
      some software, results of own experiments or conceptual
      work, or especially interesting information from
      somewhere else.

      Reports should be clearly indicated to be such by giving
      them a subject line of the form "Report:
      this-and-that"

    7. Discussions
    ++++++++++++++

      An especially valuable possibility of Usenet is of course
      that of discussing a certain topic with hundreds of potential
      participants. All traffic in the newsgroup that can not be
      subsumed under one of the above categories should belong
      to a discussion.

      If somebody explicitly wants to start a discussion, he/she
      can do so by giving the posting a subject line of the form
      "Subject: Discussion: this-and-that"

      It is quite difficult to keep a discussion from drifting into
      chaos, but, unfortunately, as many many other newsgroups
      show there seems to be no secure way to avoid this. On the
      other hand, comp.ai.neural-nets has not had many
      problems with this effect in the past, so let's just go and
      hope...

   ------------------------------------------------------------------------

 2. A: What is a neural network (NN)?
 ====================================

   First of all, when we are talking about a neural network, we
   *should* usually better say "artificial neural network" (ANN),
   because that is what we mean most of the time. Biological neural
   networks are much more complicated in their elementary
   structures than the mathematical models we use for ANNs.

   A vague description is as follows:

   An ANN is a network of many very simple processors ("units"),
   each possibly having a (small amount of) local memory. The units
   are connected by unidirectional communication channels
   ("connections"), which carry numeric (as opposed to symbolic)
   data. The units operate only on their local data and on the inputs
   they receive via the connections.

   The design motivation is what distinguishes neural networks from
   other mathematical techniques:

   A neural network is a processing device, either an algorithm, or
   actual hardware, whose design was motivated by the design and
   functioning of human brains and components thereof.

   Most neural networks have some sort of "training" rule whereby
   the weights of connections are adjusted on the basis of presented
   patterns. In other words, neural networks "learn" from examples,
   just like children learn to recognize dogs from examples of dogs,
   and exhibit some structural capability for generalization.

   Neural networks normally have great potential for parallelism,
   since the computations of the components are independent of each
   other.

   ------------------------------------------------------------------------

 3. A: What can you do with a Neural Network and
 ===============================================
   what not?
   =========

   In principle, NNs can compute any computable function, i.e. they
   can do everything a normal digital computer can do. Especially
   anything that can be represented as a mapping between vector
   spaces can be approximated to arbitrary precision by feedforward
   NNs (which is the most often used type).

   In practice, NNs are especially useful for mapping problems which
   are tolerant of some errors, have lots of example data available,
   but to which hard and fast rules can not easily be applied. NNs
   are, at least today, difficult to apply successfully to problems that
   concern manipulation of symbols and memory.

   ------------------------------------------------------------------------

 4. A: Who is concerned with Neural Networks?
 ============================================

   Neural Networks are interesting for quite a lot of very dissimilar
   people:
    o Computer scientists want to find out about the properties
      of non-symbolic information processing with neural nets
      and about learning systems in general.
    o Engineers of many kinds want to exploit the capabilities of
      neural networks on many areas (e.g. signal processing) to
      solve their application problems.
    o Cognitive scientists view neural networks as a possible
      apparatus to describe models of thinking and conscience
      (High-level brain function).
    o Neuro-physiologists use neural networks to describe and
      explore medium-level brain function (e.g. memory, sensory
      system, motorics).
    o Physicists use neural networks to model phenomena in
      statistical mechanics and for a lot of other tasks.
    o Biologists use Neural Networks to interpret nucleotide
      sequences.
    o Philosophers and some other people may also be interested
      in Neural Networks for various reasons.

   ------------------------------------------------------------------------

 5. A: What does 'backprop' mean? What is
 ========================================
   'overfitting'?
   ===============

   'Backprop' is an abbreviation for 'backpropagation of error' which
   is the most widely used learning method for neural networks
   today. Although it has many disadvantages, which could be
   summarized in the sentence "You are almost not knowing what
   you are actually doing when using backpropagation" :-) it has
   pretty much success on practical applications and is relatively easy
   to apply.

   It is for the training of layered (i.e., nodes are grouped in layers)
   feedforward (i.e., the arcs joining nodes are unidirectional, and
   there are no cycles) nets (often called "multi layer perceptrons").

   Back-propagation needs a teacher that knows the correct output
   for any input ("supervised learning") and uses gradient descent on
   the error (as provided by the teacher) to train the weights. The
   activation function is (usually) a sigmoidal (i.e., bounded above
   and below, but differentiable) function of a weighted sum of the
   nodes inputs.

   The use of a gradient descent algorithm to train its weights makes
   it slow to train; but being a feedforward algorithm, it is quite rapid
   during the recall phase.

   Literature:
      Rumelhart, D. E. and McClelland, J. L. (1986): Parallel
      Distributed Processing: Explorations in the Microstructure
      of Cognition (volume 1, pp 318-362). The MIT Press.

   (this is the classic one) or one of the dozens of other books or
   articles on backpropagation (see also answer "books").

   'Overfitting' (often also called 'overtraining' or 'overlearning') is
   the phenomenon that in most cases a network gets worse instead
   of better after a certain point during training when it is trained to
   as low errors as possible. This is because such long training may
   make the network 'memorize' the training patterns, including all
   of their peculiarities. However, one is usually interested in the
   generalization of the network, i.e., the error it exhibits on examples
   NOT seen during training. Learning the peculiarities of the
   training set makes the generalization worse. The network should
   only learn the general structure of the examples.

   There are various methods to fight overfitting. The two most
   important classes of such methods are regularization methods
   (such as weight decay) and early stopping. Regularization
   methods try to limit the complexity of the network such that it is
   unable to learn peculiarities. Early stopping aims at stopping the
   training at the point of optimal generalization. A description of the
   early stopping method can for instance be found in section 3.3 of
   /pub/papers/techreports/1994-21.ps.Z on ftp.ira.uka.de
   (anonymous ftp).

   ------------------------------------------------------------------------

 6. A: Why use a bias input? Why activation
 ==========================================
   functions?
   ===========

   One way of looking at the need for bias inputs is that the inputs to
   each unit in the net define an N-dimensional space, and the unit
   draws a hyperplane through that space, producing an "on" output
   on one side and an "off" output on the other. (With sigmoid units
   the plane will not be sharp -- there will be some gray area of
   intermediate values near the separating plane -- but ignore this
   for now.)
   The weights determine where this hyperplane is in the input space.
   Without a bias input, this separating plane is constrained to pass
   through the origin of the hyperspace defined by the inputs. For
   some problems that's OK, but in many problems the plane would
   be much more useful somewhere else. If you have many units in a
   layer, they share the same input space and without bias would
   ALL be constrained to pass through the origin.

   Activation functions are needed to introduce nonlinearity into the
   network. Without nonlinearity, hidden units would not make nets
   more powerful than just plain perceptrons (which do not have any
   hidden units, just input and output units). The reason is that a
   composition of linear functions is again a linear function.
   However, it is just the nonlinearity (i.e, the capability to represent
   nonlinear functions) that makes multilayer networks so powerful.
   Almost any nonlinear function does the job, although for
   backpropagation learning it must be differentiable and it helps if
   the function is bounded; the popular sigmoidal functions and
   gaussian functions are the most common choices.

   ------------------------------------------------------------------------

 7. A: How many hidden units should I use?
 ==========================================

   There is no way to determine a good network topology just from
   the number of inputs and outputs. It depends critically on the
   number of training examples and the complexity of the
   classification you are trying to learn. There are problems with one
   input and one output that require millions of hidden units, and
   problems with a million inputs and a million outputs that require
   only one hidden unit, or none at all.
   Some books and articles offer "rules of thumb" for choosing a
   topopology -- Ninputs plus Noutputs dividied by two, maybe with
   a square root in there somewhere -- but such rules are total
   garbage. Other rules relate to the number of examples available:
   Use at most so many hidden units that the number of weights in
   the network times 10 is smaller than the number of examples.
   Such rules are only concerned with overfitting and are unreliable
   as well.

   ------------------------------------------------------------------------

 8. A: How many learning methods for NNs exist?
 ==============================================
   Which?
   ======

   There are many many learning methods for NNs by now. Nobody
   knows exactly how many. New ones (at least variations of existing
   ones) are invented every week. Below is a collection of some of the
   most well known methods; not claiming to be complete.

   The main categorization of these methods is the distinction of
   supervised from unsupervised learning:

   In supervised learning, there is a "teacher" who in the learning
   phase "tells" the net how well it performs ("reinforcement
   learning") or what the correct behavior would have been ("fully
   supervised learning").

   In unsupervised learning the net is autonomous: it just looks at the
   data it is presented with, finds out about some of the properties of
   the data set and learns to reflect these properties in its output.
   What exactly these properties are, that the network can learn to
   recognise, depends on the particular network model and learning
   method.

   Many of these learning methods are closely connected with a
   certain (class of) network topology.

   Now here is the list, just giving some names:

   1. UNSUPERVISED LEARNING (i.e. without a "teacher"):
        1). Feedback Nets:
           a). Additive Grossberg (AG)
           b). Shunting Grossberg (SG)
           c). Binary Adaptive Resonance Theory (ART1)
           d). Analog Adaptive Resonance Theory (ART2, ART2a)
           e). Discrete Hopfield (DH)
           f). Continuous Hopfield (CH)
           g). Discrete Bidirectional Associative Memory (BAM)
           h). Temporal Associative Memory (TAM)
           i). Adaptive Bidirectional Associative Memory (ABAM)
           j). Kohonen Self-organizing Map/Topology-preserving map (SOM/TPM)
           k). Competitive learning
        2). Feedforward-only Nets:
           a). Learning Matrix (LM)
           b). Driver-Reinforcement Learning (DR)
           c). Linear Associative Memory (LAM)
           d). Optimal Linear Associative Memory (OLAM)
           e). Sparse Distributed Associative Memory (SDM)
           f). Fuzzy Associative Memory (FAM)
           g). Counterprogation (CPN)

   2. SUPERVISED LEARNING (i.e. with a "teacher"):
        1). Feedback Nets:
           a). Brain-State-in-a-Box (BSB)
           b). Fuzzy Congitive Map (FCM)
           c). Boltzmann Machine (BM)
           d). Mean Field Annealing (MFT)
           e). Recurrent Cascade Correlation (RCC)
           f). Learning Vector Quantization (LVQ)
           g). Backpropagation through time (BPTT)
           h). Real-time recurrent learning (RTRL)
           i). Recurrent Extended Kalman Filter (EKF)
        2). Feedforward-only Nets:
           a). Perceptron
           b). Adaline, Madaline
           c). Backpropagation (BP)
           d). Cauchy Machine (CM)
           e). Adaptive Heuristic Critic (AHC)
           f). Time Delay Neural Network (TDNN)
           g). Associative Reward Penalty (ARP)
           h). Avalanche Matched Filter (AMF)
           i). Backpercolation (Perc)
           j). Artmap
           k). Adaptive Logic Network (ALN)
           l). Cascade Correlation (CasCor)
           m). Extended Kalman Filter(EKF)

   ------------------------------------------------------------------------

 9. A: What about Genetic Algorithms?
 ====================================

   There are a number of definitions of GA (Genetic Algorithm). A
   possible one is

     A GA is an optimization program
     that starts with
     a population of encoded procedures,       (Creation of Life :-> )
     mutates them stochastically,              (Get cancer or so :-> )
     and uses a selection process              (Darwinism)
     to prefer the mutants with high fitness
     and perhaps a recombination process       (Make babies :-> )
     to combine properties of (preferably) the succesful mutants.

   Genetic Algorithms are just a special case of the more general idea
   of ``evolutionary computation''. There is a newsgroup that is
   dedicated to the field of evolutionary computation called
   comp.ai.genetic. It has a detailed FAQ posting which, for instance,
   explains the terms "Genetic Algorithm", "Evolutionary
   Programming", "Evolution Strategy", "Classifier System", and
   "Genetic Programming". That FAQ also contains lots of pointers
   to relevant literature, software, other sources of information, et
   cetera et cetera. Please see the comp.ai.genetic FAQ for further
   information.

   ------------------------------------------------------------------------

10. A: What about Fuzzy Logic?
==============================

   Fuzzy Logic is an area of research based on the work of L.A.
   Zadeh. It is a departure from classical two-valued sets and logic,
   that uses "soft" linguistic (e.g. large, hot, tall) system variables and
   a continuous range of truth values in the interval [0,1], rather
   than strict binary (True or False) decisions and assignments.

   Fuzzy logic is used where a system is difficult to model exactly
   (but an inexact model is available), is controlled by a human
   operator or expert, or where ambiguity or vagueness is common. A
   typical fuzzy system consists of a rule base, membership functions,
   and an inference procedure.

   Most Fuzzy Logic discussion takes place in the newsgroup
   comp.ai.fuzzy, but there is also some work (and discussion) about
   combining fuzzy logic with Neural Network approaches in
   comp.ai.neural-nets.

   For more details see (for example):

   Klir, G.J. and Folger, T.A.: Fuzzy Sets, Uncertainty, and
   Information Prentice-Hall, Englewood Cliffs, N.J., 1988.
   Kosko, B.: Neural Networks and Fuzzy Systems Prentice Hall,
   Englewood Cliffs, NJ, 1992.

   ------------------------------------------------------------------------

11. A: How are NNs related to statistical methods?
===================================================

   There is considerable overlap between the fields of neural
   networks and statistics.
   Statistics is concerned with data analysis. In neural network
   terminology, statistical inference means learning to generalize
   from noisy data. Some neural networks are not concerned with
   data analysis (e.g., those intended to model biological systems) and
   therefore have little to do with statistics. Some neural networks do
   not learn (e.g., Hopfield nets) and therefore have little to do with
   statistics. Some neural networks can learn successfully only from
   noise-free data (e.g., ART or the perceptron rule) and therefore
   would not be considered statistical methods. But most neural
   networks that can learn to generalize effectively from noisy data
   are similar or identical to statistical methods. For example:
    o Feedforward nets with no hidden layer (including
      functional-link neural nets and higher-order neural nets)
      are basically generalized linear models.
    o Feedforward nets with one hidden layer are closely related
      to projection pursuit regression.
    o Probabilistic neural nets are identical to kernel
      discriminant analysis.
    o Kohonen nets for adaptive vector quantization are very
      similar to k-means cluster analysis.
    o Hebbian learning is closely related to principal component
      analysis.
   Some neural network areas that appear to have no close relatives
   in the existing statistical literature are:
    o Kohonen's self-organizing maps.
    o Reinforcement learning ((although this is treated in the
      operations research literature as Markov decision
      processes).
    o Stopped training (the purpose and effect of stopped training
      are similar to shrinkage estimation, but the method is quite
      different).
   Feedforward nets are a subset of the class of nonlinear regression
   and discrimination models. Statisticians have studied the
   properties of this general class but had not considered the specific
   case of feedforward neural nets before such networks were
   popularized in the neural network field. Still, many results from
   the statistical theory of nonlinear models apply directly to
   feedforward nets, and the methods that are commonly used for
   fitting nonlinear models, such as various Levenberg-Marquardt
   and conjugate gradient algorithms, can be used to train
   feedforward nets.

   While neural nets are often defined in terms of their algorithms or
   implementations, statistical methods are usually defined in terms
   of their results. The arithmetic mean, for example, can be
   computed by a (very simple) backprop net, by applying the usual
   formula SUM(x_i)/n, or by various other methods. What you get
   is still an arithmetic mean regardless of how you compute it. So a
   statistician would consider standard backprop, Quickprop, and
   Levenberg-Marquardt as different algorithms for implementing
   the same statistical model such as a feedforward net. On the other
   hand, different training criteria, such as least squares and cross
   entropy, are viewed by statisticians as fundamentally different
   estimation methods with different statistical properties.

   It is sometimes claimed that neural networks, unlike statistical
   models, require no distributional assumptions. In fact, neural
   networks involve exactly the same sort of distributional
   assumptions as statistical models, but statisticians study the
   consequences and importance of these assumptions while most
   neural networkers ignore them. For example, least-squares
   training methods are widely used by statisticians and neural
   networkers. Statisticians realize that least-squares training
   involves implicit distributional assumptions in that least-squares
   estimates have certain optimality properties for noise that is
   normally distributed with equal variance for all training cases and
   that is independent between different cases. These optimality
   properties are consequences of the fact that least-squares
   estimation is maximum likelihood under those conditions.
   Similarly, cross-entropy is maximum likelihood for noise with a
   Bernoulli distribution. If you study the distributional assumptions,
   then you can recognize and deal with violations of the
   assumptions. For example, if you have normally distributed noise
   but some training cases have greater noise variance than others,
   then you may be able to use weighted least squares instead of
   ordinary least squares to obtain more efficient estimates.

   Here are a few references:

   Chatfield, C. (1993), "Neural networks: Forecasting breakthrough
   or passing fad", International Journal of Forecasting, 9, 1-3.

   Cheng, B. and Titterington, D.M. (1994), "Neural Networks: A
   Review from a Statistical Perspective", Statistical Science, 9,
   2-54.

   Geman, S., Bienenstock, E. and Doursat, R. (1992), "Neural
   Networks and the Bias/Variance Dilemma", Neural Computation,
   4, 1-58.

   Kushner, H. & Clark, D. (1978), _Stochastic Approximation
   Methods for Constrained and Unconstrained Systems_,
   Springer-Verlag.

   Michie, D., Spiegelhalter, D.J. and Taylor, C.C. (1994), _Machine
   Learning, Neural and Statistical Classification_, Ellis Horwood.

   Ripley, B.D. (1993), "Statistical Aspects of Neural Networks", in
   O.E. Barndorff-Nielsen, J.L. Jensen and W.S. Kendall, eds.,
   _Networks and Chaos: Statistical and Probabilistic Aspects_,
   Chapman & Hall. ISBN 0 412 46530 2.

   Sarle, W.S. (1994), "Neural Networks and Statistical Models,"
   Proceedings of the Nineteenth Annual SAS Users Group
   International Conference, Cary, NC: SAS Institute, pp 1538-1550.
   ( ftp://ftp.sas.com/pub/sugi19/neural/neural1.ps)

   White, H. (1989), "Learning in Artificial Neural Networks: A
   Statistical Perspective," Neural Computation, 1, 425-464.

   White, H. (1992), _Artificial Neural Networks: Approximation
   and Learning Theory_, Blackwell.

   ------------------------------------------------------------------------

12. A: Good introductory literature about Neural
================================================
   Networks?
   =========

   0.) The best (subjectively, of course -- please don't flame me):
   ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

   Haykin, S. (1994). Neural Networks, a Comprehensive
   Foundation. Macmillan, New York, NY. "A very readable, well
   written intermediate to advanced text on NNs Perspective is
   primarily one of pattern recognition, estimation and signal
   processing. However, there are well-written chapters on
   neurodynamics and VLSI implementation. Though there is
   emphasis on formal mathematical models of NNs as universal
   approximators, statistical estimators, etc., there are also examples
   of NNs used in practical applications. The problem sets at the end
   of each chapter nicely complement the material. In the
   bibliography are over 1000 references. If one buys only one book
   on neural networks, this should be it."

   Hertz, J., Krogh, A., and Palmer, R. (1991). Introduction to the
   Theory of Neural Computation. Addison-Wesley: Redwood City,
   California. ISBN 0-201-50395-6 (hardbound) and
   0-201-51560-1 (paperbound) Comments: "My first impression is
   that this one is by far the best book on the topic. And it's below
   $30 for the paperback."; "Well written, theoretical (but not
   overwhelming)"; It provides a good balance of model development,
   computational algorithms, and applications. The mathematical
   derivations are especially well done"; "Nice mathematical analysis
   on the mechanism of different learning algorithms"; "It is NOT
   for mathematical beginner. If you don't have a good grasp of
   higher level math, this book can be really tough to get through."

   Masters,Timothy (1994). Practical Neural Network Recipes in
   C++. Academic Press, ISBN 0-12-479040-2, US $45 incl. disks.
   "Lots of very good practical advice which most other books lack."

   1.) Books for the beginner:
   +++++++++++++++++++++++++++

   Aleksander, I. and Morton, H. (1990). An Introduction to Neural
   Computing. Chapman and Hall. (ISBN 0-412-37780-2).
   Comments: "This book seems to be intended for the first year of
   university education."

   Beale, R. and Jackson, T. (1990). Neural Computing, an
   Introduction. Adam Hilger, IOP Publishing Ltd : Bristol. (ISBN
   0-85274-262-2). Comments: "It's clearly written. Lots of hints as
   to how to get the adaptive models covered to work (not always
   well explained in the original sources). Consistent mathematical
   terminology. Covers perceptrons, error-backpropagation, Kohonen
   self-org model, Hopfield type models, ART, and associative
   memories."

   Dayhoff, J. E. (1990). Neural Network Architectures: An
   Introduction. Van Nostrand Reinhold: New York. Comments:
   "Like Wasserman's book, Dayhoff's book is also very easy to
   understand".

   Fausett, L. V. (1994). Fundamentals of Neural Networks:
   Architectures, Algorithms and Applications, Prentice Hall, ISBN
   0-13-334186-0. Also published as a Prentice Hall International
   Edition, ISBN 0-13-042250-9. Sample softeware (source code
   listings in C and Fortran) is included in an Instructor's Manual.
   "Intermediate in level between Wasserman and
   Hertz/Krogh/Palmer. Algorithms for a broad range of neural
   networks, including a chapter on Adaptive Resonace Theory with
   ART2. Simple examples for each network."

   Freeman, James (1994). Simulating Neural Networks with
   Mathematica, Addison-Wesley, ISBN: 0-201-56629-X. Helps
   the reader make his own NNs. The mathematica code for the
   programs in the book is also available through the internet: Send
   mail to [e-mail] or try http://www.wri.com/ on the
   World Wide Web.

   Hecht-Nielsen, R. (1990). Neurocomputing. Addison Wesley.
   Comments: "A good book", "comprises a nice historical overview
   and a chapter about NN hardware. Well structured prose. Makes
   important concepts clear."

   McClelland, J. L. and Rumelhart, D. E. (1988). Explorations in
   Parallel Distributed Processing: Computational Models of
   Cognition and Perception (software manual). The MIT Press.
   Comments: "Written in a tutorial style, and includes 2 diskettes of
   NN simulation programs that can be compiled on MS-DOS or
   Unix (and they do too !)"; "The programs are pretty reasonable as
   an introduction to some of the things that NNs can do."; "There
   are *two* editions of this book. One comes with disks for the IBM
   PC, the other comes with disks for the Macintosh".

   McCord Nelson, M. and Illingworth, W.T. (1990). A Practical
   Guide to Neural Nets. Addison-Wesley Publishing Company, Inc.
   (ISBN 0-201-52376-0). Comments: "No formulas at all"; "It
   does not have much detailed model development (very few
   equations), but it does present many areas of application. It
   includes a chapter on current areas of research. A variety of
   commercial applications is discussed in chapter 1. It also includes a
   program diskette with a fancy graphical interface (unlike the PDP
   diskette)".

   Muller, B. and Reinhardt, J. (1990). Neural Networks, An
   Introduction. Springer-Verlag: Berlin Heidelberg New York
   (ISBN: 3-540-52380-4 and 0-387-52380-4). Comments: The
   book was developed out of a course on neural-network models
   with computer demonstrations that was taught by the authors to
   Physics students. The book comes together with a PC-diskette.
   The book is divided into three parts: (1) Models of Neural
   Networks; describing several architectures and learing rules,
   including the mathematics. (2) Statistical Physiscs of Neural
   Networks; "hard-core" physics section developing formal theories
   of stochastic neural networks. (3) Computer Codes; explanation
   about the demonstration programs. First part gives a nice
   introduction into neural networks together with the formulas.
   Together with the demonstration programs a 'feel' for neural
   networks can be developed.

   Orchard, G.A. & Phillips, W.A. (1991). Neural Computation: A
   Beginner's Guide. Lawrence Earlbaum Associates: London.
   Comments: "Short user-friendly introduction to the area, with a
   non-technical flavour. Apparently accompanies a software
   package, but I haven't seen that yet".

   Rao, V.B & H.V. (1993). C++ Neural Networks and Fuzzy Logic.
   MIS:Press, ISBN 1-55828-298-x, US $45 incl. disks. "Probably
   not 'leading edge' stuff but detailed enough to get your hands
   dirty!"

   Wasserman, P. D. (1989). Neural Computing: Theory & Practice.
   Van Nostrand Reinhold: New York. (ISBN 0-442-20743-3)
   Comments: "Wasserman flatly enumerates some common
   architectures from an engineer's perspective ('how it works')
   without ever addressing the underlying fundamentals ('why it
   works') - important basic concepts such as clustering, principal
   components or gradient descent are not treated. It's also full of
   errors, and unhelpful diagrams drawn with what appears to be
   PCB board layout software from the '70s. For anyone who wants
   to do active research in the field I consider it quite inadequate";
   "Okay, but too shallow"; "Quite easy to understand"; "The best
   bedtime reading for Neural Networks. I have given this book to
   numerous collegues who want to know NN basics, but who never
   plan to implement anything. An excellent book to give your
   manager."

   Wasserman, P.D. (1993). Advanced Methods in Neural
   Computing. Van Nostrand Reinhold: New York (ISBN:
   0-442-00461-3). Comments: Several neural network topics are
   discussed e.g. Probalistic Neural Networks, Backpropagation and
   beyond, neural control, Radial Basis Function Networks, Neural
   Engineering. Furthermore, several subjects related to neural
   networks are mentioned e.g. genetic algorithms, fuzzy logic, chaos.
   Just the functionality of these subjects is described; enough to get
   you started. Lots of references are given to more elaborate
   descriptions. Easy to read, no extensive mathematical background
   necessary.

   2.) The classics:
   +++++++++++++++++

   Kohonen, T. (1984). Self-organization and Associative Memory.
   Springer-Verlag: New York. (2nd Edition: 1988; 3rd edition:
   1989). Comments: "The section on Pattern mathematics is
   excellent."

   Rumelhart, D. E. and McClelland, J. L. (1986). Parallel
   Distributed Processing: Explorations in the Microstructure of
   Cognition (volumes 1 & 2). The MIT Press. Comments: "As a
   computer scientist I found the two Rumelhart and McClelland
   books really heavy going and definitely not the sort of thing to
   read if you are a beginner."; "It's quite readable, and affordable
   (about $65 for both volumes)."; "THE Connectionist bible".

   3.) Introductory journal articles:
   ++++++++++++++++++++++++++++++++++

   Hinton, G. E. (1989). Connectionist learning procedures. Artificial
   Intelligence, Vol. 40, pp. 185--234. Comments: "One of the better
   neural networks overview papers, although the distinction
   between network topology and learning algorithm is not always
   very clear. Could very well be used as an introduction to neural
   networks."

   Knight, K. (1990). Connectionist, Ideas and Algorithms.
   Communications of the ACM. November 1990. Vol.33 nr.11, pp
   59-74. Comments:"A good article, while it is for most people easy
   to find a copy of this journal."

   Kohonen, T. (1988). An Introduction to Neural Computing.
   Neural Networks, vol. 1, no. 1. pp. 3-16. Comments: "A general
   review".

   4.) Not-quite-so-introductory literature:
   +++++++++++++++++++++++++++++++++++++++++

   Anderson, J. A. and Rosenfeld, E. (Eds). (1988). Neurocomputing:
   Foundations of Research. The MIT Press: Cambridge, MA.
   Comments: "An expensive book, but excellent for reference. It is a
   collection of reprints of most of the major papers in the field."

   Anderson, J. A., Pellionisz, A. and Rosenfeld, E. (Eds). (1990).
   Neurocomputing 2: Directions for Research. The MIT Press:
   Cambridge, MA. Comments: "The sequel to their well-known
   Neurocomputing book."

   Caudill, M. and Butler, C. (1990). Naturally Intelligent Systems.
   MIT Press: Cambridge, Massachusetts. (ISBN 0-262-03156-6).
   Comments: "I guess one of the best books I read"; "May not be
   suited for people who want to do some research in the area".

   Cichocki, A. and Unbehauen, R. (1994). Neural Networks for
   Optimization and Signal Processing. John Wiley & Sons, West
   Sussex, England, 1993, ISBN 0-471-930105 (hardbound), 526
   pages, $57.95. "Partly a textbook and partly a research
   monograph; introduces the basic concepts, techniques, and models
   related to neural networks and optimization, excluding rigorous
   mathematical details. Accessible to a wide readership with a
   differential calculus background. The main coverage of the book is
   on recurrent neural networks with continuous state variables. The
   book title would be more appropriate without mentioning signal
   processing. Well edited, good illustrations."

   Khanna, T. (1990). Foundations of Neural Networks.
   Addison-Wesley: New York. Comments: "Not so bad (with a
   page of erroneous formulas (if I remember well), and #hidden
   layers isn't well described)."; "Khanna's intention in writing his
   book with math analysis should be commended but he made
   several mistakes in the math part".

   Kung, S.Y. (1993). Digital Neural Networks, Prentice Hall,
   Englewood Cliffs, NJ.

   Levine, D. S. (1990). Introduction to Neural and Cognitive
   Modeling. Lawrence Erlbaum: Hillsdale, N.J. Comments: "Highly
   recommended".

   Lippmann, R. P. (April 1987). An introduction to computing with
   neural nets. IEEE Acoustics, Speech, and Signal Processing
   Magazine. vol. 2, no. 4, pp 4-22. Comments: "Much acclaimed as
   an overview of neural networks, but rather inaccurate on several
   points. The categorization into binary and continuous- valued
   input neural networks is rather arbitrary, and may work confusing
   for the unexperienced reader. Not all networks discussed are of
   equal importance."

   Maren, A., Harston, C. and Pap, R., (1990). Handbook of Neural
   Computing Applications. Academic Press. ISBN: 0-12-471260-6.
   (451 pages) Comments: "They cover a broad area"; "Introductory
   with suggested applications implementation".

   Pao, Y. H. (1989). Adaptive Pattern Recognition and Neural
   Networks Addison-Wesley Publishing Company, Inc. (ISBN
   0-201-12584-6) Comments: "An excellent book that ties together
   classical approaches to pattern recognition with Neural Nets. Most
   other NN books do not even mention conventional approaches."

   Rumelhart, D. E., Hinton, G. E. and Williams, R. J. (1986).
   Learning representations by back-propagating errors. Nature, vol
   323 (9 October), pp. 533-536. Comments: "Gives a very good
   potted explanation of backprop NN's. It gives sufficient detail to
   write your own NN simulation."

   Simpson, P. K. (1990). Artificial Neural Systems: Foundations,
   Paradigms, Applications and Implementations. Pergamon Press:
   New York. Comments: "Contains a very useful 37 page
   bibliography. A large number of paradigms are presented. On the
   negative side the book is very shallow. Best used as a complement
   to other books".

   Zeidenberg. M. (1990). Neural Networks in Artificial Intelligence.
   Ellis Horwood, Ltd., Chichester. Comments: "Gives the AI point
   of view".

   Zornetzer, S. F., Davis, J. L. and Lau, C. (1990). An Introduction
   to Neural and Electronic Networks. Academic Press. (ISBN
   0-12-781881-2) Comments: "Covers quite a broad range of
   topics (collection of articles/papers )."; "Provides a primer-like
   introduction and overview for a broad audience, and employs a
   strong interdisciplinary emphasis".

   ------------------------------------------------------------------------

13. A: Any journals and magazines about Neural
==============================================
   Networks?
   =========

   [to be added: comments on speed of reviewing and publishing,
                 whether they accept TeX format or ASCII by e-mail, etc.]

   A. Dedicated Neural Network Journals:
   +++++++++++++++++++++++++++++++++++++

   Title:   Neural Networks
   Publish: Pergamon Press
   Address: Pergamon Journals Inc., Fairview Park, Elmsford,
            New York 10523, USA and Pergamon Journals Ltd.
            Headington Hill Hall, Oxford OX3, 0BW, England
   Freq.:   10 issues/year (vol. 1 in 1988)
   Cost/Yr: Free with INNS or JNNS or ENNS membership ($45?),
            Individual $65, Institution $175
   ISSN #:  0893-6080
   Remark:  Official Journal of International Neural Network Society (INNS),
            European Neural Network Society (ENNS) and Japanese Neural
            Network Society (JNNS).
            Contains Original Contributions, Invited Review Articles, Letters
            to Editor, Book Reviews, Editorials, Announcements, Software Surveys.

   Title:   Neural Computation
   Publish: MIT Press
   Address: MIT Press Journals, 55 Hayward Street Cambridge,
            MA 02142-9949, USA, Phone: [phone]
   Freq.:   Quarterly (vol. 1 in 1989)
   Cost/Yr: Individual $45, Institution $90, Students $35; Add $9 Outside USA
   ISSN #:  0899-7667
   Remark:  Combination of Reviews (10,000 words), Views (4,000 words)
            and Letters (2,000 words).  I have found this journal to be of
            outstanding quality.
            (Note: Remarks supplied by Mike Plonski "[e-mail]")

   Title:   IEEE Transactions on Neural Networks
   Publish: Institute of Electrical and Electronics Engineers (IEEE)
   Address: IEEE Service Cemter, 445 Hoes Lane, P.O. Box 1331, Piscataway, NJ,
            08855-1331 USA. Tel: [phone]
   Cost/Yr: $10 for Members belonging to participating IEEE societies
   Freq.:   Quarterly (vol. 1 in March 1990)
   Remark:  Devoted to the science and technology of neural networks
            which disclose significant  technical knowledge, exploratory
            developments and applications of neural networks from biology to
            software to hardware.  Emphasis is on artificial neural networks.
            Specific aspects include self organizing systems, neurobiological
            connections, network dynamics and architecture, speech recognition,
            electronic and photonic implementation, robotics and controls.
            Includes Letters concerning new research results.
            (Note: Remarks are from journal announcement)

   Title:   International Journal of Neural Systems
   Publish: World Scientific Publishing
   Address: USA: World Scientific Publishing Co., 1060 Main Street, River Edge,
            NJ 07666. Tel: [phone]; Europe: World Scientific Publishing
            Co. Ltd., 57 Shelton Street, London WC2H 9HE, England.
            Tel: [phone]; Asia: World Scientific Publishing Co. Pte. Ltd.,
            1022 Hougang Avenue 1 #05-3520, Singapore 1953, Rep. of Singapore
            Tel: [phone].
   Freq.:   Quarterly (Vol. 1 in 1990)
   Cost/Yr: Individual $122, Institution $255 (plus $15-$25 for postage)
   ISSN #:  0129-0657 (IJNS)
   Remark:  The International Journal of Neural Systems is a quarterly
            journal which covers information processing in natural
            and artificial neural systems. Contributions include research papers,
            reviews, and Letters to the Editor - communications under 3,000
            words in length, which are published within six months of receipt.
            Other contributions are typically published within nine months.
            The journal presents a fresh undogmatic attitude towards this
            multidisciplinary field and aims to be a forum for novel ideas and
            improved understanding of collective and cooperative phenomena with
            computational capabilities.
            Papers should be submitted to World Scientific's UK office. Once a
            paper is accepted for publication, authors are invited to e-mail
            the LaTeX source file of their paper in order to expedite publication.

   Title:   International Journal of Neurocomputing
   Publish: Elsevier Science Publishers, Journal Dept.; PO Box 211;
            1000 AE Amsterdam, The Netherlands
   Freq.:   Quarterly (vol. 1 in 1989)
   Editor:  V.D. Sanchez A.; German Aerospace Research Establishment;
            Institute for Robotics and System Dynamics, 82230 Wessling, Germany.
            Current events and software news editor: Dr. F. Murtagh, ESA,
            Karl-Schwarzschild Strasse 2, D-85748, Garching, Germany,
            phone [phone], fax [phone], email [e-mail]

   Title:   Neural Processing Letters
   Publish: D facto publications
   Address: 45 rue Masui; B-1210 Brussels, Belgium
            Phone: [phone];  Fax: [phone]
   Freq:    6 issues/year (vol. 1 in September 1994)
   Cost/Yr: BEF 4400 (about $140)
   ISSN #:  1370-4621
   Remark:  The aim of the journal is to rapidly publish new ideas, original
            developments and work in progress.  Neural Processing Letters
            covers all aspects of the Artificial Neural Networks field.
            Publication delay is about 3 months.
            FTP server available:
             ftp://ftp.dice.ucl.ac.be/pub/neural-nets/NPL.
            WWW server available:
               http://www.dice.ucl.ac.be/neural-nets/NPL/NPL.html

   Title:   Neural Network News
   Publish: AIWeek Inc.
   Address: Neural Network News, 2555 Cumberland Parkway, Suite 299,
            Atlanta, GA 30339 USA. Tel: [phone]
   Freq.:   Monthly (beginning September 1989)
   Cost/Yr: USA and Canada $249, Elsewhere $299
   Remark:  Commercial Newsletter

   Title:   Network: Computation in Neural Systems
   Publish: IOP Publishing Ltd
   Address: Europe: IOP Publishing Ltd, Techno House, Redcliffe Way, Bristol
            BS1 6NX, UK; IN USA: American Institute of Physics, Subscriber
            Services 500 Sunnyside Blvd., Woodbury, NY  11797-2999
   Freq.:   Quarterly (1st issue 1990)
   Cost/Yr: USA: $180,  Europe: 110 pounds
   Remark:  Description: "a forum for integrating theoretical and experimental
            findings across relevant interdisciplinary boundaries."  Contents:
            Submitted articles reviewed by two technical referees  paper's
            interdisciplinary format and accessability."  Also Viewpoints and
            Reviews commissioned by the editors, abstracts (with reviews) of
            articles published in other journals, and book reviews.
            Comment: While the price discourages me (my comments are based
            upon a free sample copy), I think that the journal succeeds
            very well.  The highest density of interesting articles I
            have found in any journal.
            (Note: Remarks supplied by [e-mail])

   Title:   Connection Science: Journal of Neural Computing,
            Artificial Intelligence and Cognitive Research
   Publish: Carfax Publishing
   Address: Europe: Carfax Publishing Company, P. O. Box 25, Abingdon,
            Oxfordshire  OX14 3UE, UK.  USA: Carafax Publishing Company,
            85 Ash Street, Hopkinton, MA 01748
   Freq.:   Quarterly (vol. 1 in 1989)
   Cost/Yr: Individual $82, Institution $184, Institution (U.K.) 74 pounds

   Title:   International Journal of Neural Networks
   Publish: Learned Information
   Freq.:   Quarterly (vol. 1 in 1989)
   Cost/Yr: 90 pounds
   ISSN #:  0954-9889
   Remark:  The journal contains articles, a conference report (at least the
            issue I have), news and a calendar.
            (Note: remark provided by J.R.M. Smits "[e-mail]")

   Title:   Sixth Generation Systems (formerly Neurocomputers)
   Publish: Gallifrey Publishing
   Address: Gallifrey Publishing, PO Box 155, Vicksburg, Michigan, 49097, USA
            Tel: [phone] fax
   Freq.    Monthly (1st issue January, 1987)
   ISSN #:  0893-1585
   Editor:  Derek F. Stubbs
   Cost/Yr: $79 (USA, Canada), US$95 (elsewhere)
   Remark:  Runs eight to 16 pages monthly. In 1995 will go to floppy disc-based
   publishing with databases +, "the equivalent to 50 pages per issue are
   planned." Often focuses on specific topics: e.g., August, 1994 contains two
   articles: "Economics, Times Series and the Market," and "Finite Particle
   Analysis - [part] II."  Stubbs also directs the company Advanced Forecasting
   Technologies. (Remark by Ed Rosenfeld: [e-mail])

   Title:   JNNS Newsletter (Newsletter of the Japan Neural Network Society)
   Publish: The Japan Neural Network Society
   Freq.:   Quarterly (vol. 1 in 1989)
   Remark:  (IN JAPANESE LANGUAGE) Official Newsletter of the Japan Neural
            Network Society(JNNS)
            (Note: remarks by Osamu Saito "[e-mail]")

   Title:   Neural Networks Today
   Remark:  I found this title in a bulletin board of october last year.
            It was a message of Tim Pattison, [e-mail]
            (Note: remark provided by J.R.M. Smits "[e-mail]")

   Title:   Computer Simulations in Brain Science

   Title:   Internation Journal of Neuroscience

   Title:   Neural Network Computation
   Remark:  Possibly the same as "Neural Computation"

   Title:   Neural Computing and Applications
   Freq.:   Quarterly
   Publish: Springer Verlag
   Cost/yr: 120 Pounds
   Remark:  Is the journal of the Neural Computing Applications Forum.
            Publishes original research and other information
            in the field of practical applications of neural computing.

   B. NN Related Journals:
   +++++++++++++++++++++++

   Title:   Complex Systems
   Publish: Complex Systems Publications
   Address: Complex Systems Publications, Inc., P.O. Box 6149, Champaign,
            IL 61821-8149, USA
   Freq.:   6 times per year (1st volume is 1987)
   ISSN #:  0891-2513
   Cost/Yr: Individual $75, Institution $225
   Remark:  Journal COMPLEX SYSTEMS  devotes to rapid publication of research
            on science, mathematics, and engineering of systems with simple
            components but complex overall behavior. Send mail to
            "[e-mail]" for additional info.
            (Remark is from announcement on Net)

   Title:   Biological Cybernetics (Kybernetik)
   Publish: Springer Verlag
   Remark:  Monthly (vol. 1 in 1961)

   Title:   Various IEEE Transactions and Magazines
   Publish: IEEE
   Remark:  Primarily see IEEE Trans. on System, Man and Cybernetics;
            Various Special Issues: April 1990 IEEE Control Systems
            Magazine.; May 1989 IEEE Trans. Circuits and Systems.;
            July 1988 IEEE Trans. Acoust. Speech Signal Process.

   Title:   The Journal of Experimental and Theoretical Artificial Intelligence
   Publish: Taylor & Francis, Ltd.
   Address: London, New York, Philadelphia
   Freq.:   ? (1st issue Jan 1989)
   Remark:  For submission information, please contact either of the editors:
            Eric Dietrich                        Chris Fields
            PACSS - Department of Philosophy     Box 30001/3CRL
            SUNY Binghamton                      New Mexico State University
            Binghamton, NY 13901                 Las Cruces, NM 88003-0001
            [e-mail]  [e-mail]

   Title:   The Behavioral and Brain Sciences
   Publish: Cambridge University Press
   Remark:  (Expensive as hell, I'm sure.)
            This is a delightful journal that encourages discussion on a
            variety of controversial topics.  I have especially enjoyed
            reading some papers in there by Dana Ballard and Stephen
            Grossberg (separate papers, not collaborations) a few years
            back.  They have a really neat concept: they get a paper,
            then invite a number of noted scientists in the field to
            praise it or trash it.  They print these commentaries, and
            give the author(s) a chance to make a rebuttal or
            concurrence.  Sometimes, as I'm sure you can imagine, things
            get pretty lively.  I'm reasonably sure they are still at
            it--I think I saw them make a call for reviewers a few
            months ago.  Their reviewers are called something like
            Behavioral and Brain Associates, and I believe they have to
            be nominated by current associates, and should be fairly
            well established in the field.  That's probably more than I
            really know about it but maybe if you post it someone who
            knows more about it will correct any errors I have made.
            The main thing is that I liked the articles I read. (Note:
            remarks by Don Wunsch )

   Title:   International Journal of Applied Intelligence
   Publish: Kluwer Academic Publishers
   Remark:  first issue in 1990(?)

   Title:   Bulletin of Mathematica Biology

   Title:   Intelligence

   Title:   Journal of Mathematical Biology

   Title:   Journal of Complex System

   Title:   AI Expert
   Publish: Miller Freeman Publishing Co., for subscription call ++415-267-7672.
   Remark:  Regularly includes ANN related articles, product
            announcements, and application reports. Listings of ANN
            programs are available on AI Expert affiliated BBS's

   Title:   International Journal of Modern Physics C
   Publish: USA: World Scientific Publishing Co., 1060 Main Street, River Edge,
            NJ 07666. Tel: [phone]; Europe: World Scientific Publishing
            Co. Ltd., 57 Shelton Street, London WC2H 9HE, England.
            Tel: [phone]; Asia: World Scientific Publishing Co. Pte. Ltd.,
            1022 Hougang Avenue 1 #05-3520, Singapore 1953, Rep. of Singapore
            Tel: [phone].
   Freq:    bi-monthly
   Eds:     H. Herrmann, R. Brower, G.C. Fox and S Nose

   Title:   Machine Learning
   Publish: Kluwer Academic Publishers
   Address: Kluwer Academic Publishers
            P.O. Box 358
            Accord Station
            Hingham, MA 02018-0358 USA
   Freq.:   Monthly (8 issues per year; increasing to 12 in 1993)
   Cost/Yr: Individual $140 (1992); Member of AAAI or CSCSI $88
   Remark:  Description: Machine Learning is an international forum for
            research on computational approaches to learning.  The journal
            publishes articles reporting substantive research results on a
            wide range of learning methods applied to a variety of task
            domains.  The ideal paper will make a theoretical contribution
            supported by a computer implementation.
            The journal has published many key papers in learning theory,
            reinforcement learning, and decision tree methods.  Recently
            it has published a special issue on connectionist approaches
            to symbolic reasoning.  The journal regularly publishes
            issues devoted to genetic algorithms as well.

   Title:   INTELLIGENCE - The Future of Computing
   Published by: Intelligence
   Address: INTELLIGENCE, P.O. Box 20008, New York, NY 10025-1510, USA,
   212-222-1123 voice & fax; email: [e-mail], CIS: 72400,1013
   Freq.    Monthly plus four special reports each year (1st issue: May, 1984)
   ISSN #:  1042-4296
   Editor:  Edward Rosenfeld
   Cost/Yr: $395 (USA), US$450 (elsewhere)
   Remark:  Has absorbed several other newsletters, like Synapse/Connection
            and Critical Technology Trends (formerly AI Trends).
            Covers NN, genetic algorithms, fuzzy systems, wavelets, chaos
            and other advanced computing approaches, as well as molecular
            computing and nanotechnology.

   Title:   Journal of Physics A: Mathematical and General
   Publish: Inst. of Physics, Bristol
   Freq:    24 issues per year.
   Remark:  Statistical mechanics aspects of neural networks
            (mostly Hopfield models).

   Title:   Physical Review A: Atomic, Molecular and Optical Physics
   Publish: The American Physical Society (Am. Inst. of Physics)
   Freq:    Monthly
   Remark:  Statistical mechanics of neural networks.

   Title:   Information Sciences
   Publish: North Holland (Elsevier Science)
   Freq.:   Monthly
   ISSN:    0020-0255
   Editor:  Paul P. Wang; Department of Electrical Engineering; Duke University;
            Durham, NC 27706, USA

   C. Journals loosely related to NNs:
   +++++++++++++++++++++++++++++++++++

   Title:   JOURNAL OF COMPLEXITY
   Remark:  (Must rank alongside Wolfram's Complex Systems)

   Title:   IEEE ASSP Magazine
   Remark:  (April 1987 had the Lippmann intro. which everyone likes to cite)

   Title:   ARTIFICIAL INTELLIGENCE
   Remark:  (Vol 40, September 1989 had the survey paper by Hinton)

   Title:   COGNITIVE SCIENCE
   Remark:  (the Boltzmann machine paper by Ackley et al appeared here
            in Vol 9, 1983)

   Title:   COGNITION
   Remark:  (Vol 28, March 1988 contained the Fodor and Pylyshyn
            critique of connectionism)

   Title:   COGNITIVE PSYCHOLOGY
   Remark:  (no comment!)

   Title:   JOURNAL OF MATHEMATICAL PSYCHOLOGY
   Remark:  (several good book reviews)

   ------------------------------------------------------------------------

14. A: The most important conferences concerned
===============================================
   with Neural Networks?
   =====================

   [to be added: has taken place how often yet; most emphasized topics;
    where to get proceedings/calls-for-papers etc. ]

   A. Dedicated Neural Network Conferences:
   ++++++++++++++++++++++++++++++++++++++++

    1. Neural Information Processing Systems (NIPS) Annually
      since 1988 in Denver, Colorado; late November or early
      December. Interdisciplinary conference with computer
      science, physics, engineering, biology, medicine, cognitive
      science topics. Covers all aspects of NNs. Proceedings
      appear several months after the conference as a book from
      Morgan Kaufman, San Mateo, CA.
    2. International Joint Conference on Neural Networks
      (IJCNN) formerly co-sponsored by INNS and IEEE, no
      longer held.
    3. Annual Conference on Neural Networks (ACNN)
    4. International Conference on Artificial Neural Networks
      (ICANN) Annually in Europe. First was 1991. Major
      conference of European Neur. Netw. Soc. (ENNS)
    5. WCNN. Sponsored by INNS.
    6. European Symposium on Artificial Neural Networks
      (ESANN). Anually since 1993 in Brussels, Belgium; late
      April; conference on the fundamental aspects of artificial
      neural networks: theory, mathematics, biology, relations
      between neural networks and other disciplines, statistics,
      learning, algorithms, models and architectures,
      self-organization, signal processing, approximation of
      functions, evolutive learning, etc. Contact: Michel
      Verleysen, D facto conference services, 45 rue Masui,
      B-1210 Brussels, Belgium, phone: [phone], fax: +
      32 2 245 46 94, e-mail: [e-mail]
    7. Artificial Neural Networks in Engineering (ANNIE)
      Anually since 1991 in St. Louis, Missouri; held in
      November. (Topics: NN architectures, pattern recognition,
      neuro-control, neuro-engineering systems. Contact:
      ANNIE; Engineering Management Department; 223
      Engineering Management Building; University of
      Missouri-Rolla; Rolla, MO 65401; FAX: [phone])
    8. many many more....

   B. Other Conferences
   ++++++++++++++++++++

    1. International Joint Conference on Artificial Intelligence
      (IJCAI)
    2. Intern. Conf. on Acustics, Speech and Signal Processing
      (ICASSP)
    3. Intern. Conf. on Pattern Recognition. Held every other
      year. Has a connectionist subconference. Information:
      General Chair Walter G. Kropatsch
      <[e-mail]>
    4. Annual Conference of the Cognitive Science Society
    5. [Vision Conferences?]

   C. Pointers to Conferences
   ++++++++++++++++++++++++++

    1. The journal "Neural Networks" has a list of conferences,
      workshops and meetings in each issue. This is quite
      interdisciplinary.
    2. There is a regular posting on comp.ai.neural-nets from
      Paultje Bakker: "Upcoming Neural Network Conferences",
      which lists names, dates, locations, contacts, and deadlines.
      It is also available for anonymous ftp from ftp.cs.uq.oz.au
      as /pub/pdp/conferences

   ------------------------------------------------------------------------

15. A: Neural Network Associations?
===================================

    1. International Neural Network Society (INNS).
    +++++++++++++++++++++++++++++++++++++++++++++++

      INNS membership includes subscription to "Neural
      Networks", the official journal of the society. Membership
      is $55 for non-students and $45 for students per year.
      Address: INNS Membership, P.O. Box 491166, Ft.
      Washington, MD 20749.

    2. International Student Society for Neural Networks
    ++++++++++++++++++++++++++++++++++++++++++++++++++++
      (ISSNNets).
      +++++++++++

      Membership is $5 per year. Address: ISSNNet, Inc., P.O.
      Box 15661, Boston, MA 02215 USA

    3. Women In Neural Network Research and technology
    ++++++++++++++++++++++++++++++++++++++++++++++++++
      (WINNERS).
      ++++++++++

      Address: WINNERS, c/o Judith Dayhoff, 11141 Georgia
      Ave., Suite 206, Wheaton, MD 20902. Phone:
      301-933-9000.

    4. European Neural Network Society (ENNS)
    +++++++++++++++++++++++++++++++++++++++++

      ENNS membership includes subscription to "Neural
      Networks", the official journal of the society. Membership
      is currently (1994) 50 UK pounds (35 UK pounds for
      students) per year. Address: ENNS Membership, Centre for
      Neural Networks, King's College London, Strand, London
      WC2R 2LS, United Kingdom.

    5. Japanese Neural Network Society (JNNS)
    +++++++++++++++++++++++++++++++++++++++++

      Address: Japanese Neural Network Society; Department of
      Engineering, Tamagawa University; 6-1-1, Tamagawa
      Gakuen, Machida City, Tokyo; 194 JAPAN; Phone: +81
      427 28 3457, Fax: [phone]

    6. Association des Connexionnistes en THese (ACTH)
    ++++++++++++++++++++++++++++++++++++++++++++++++++

      (the French Student Association for Neural Networks);
      Membership is 100 FF per year; Activities : newsletter,
      conference (every year), list of members, electronic forum;
      Journal 'Valgo' (ISSN 1243-4825); Contact : [e-mail]

    7. Neurosciences et Sciences de l'Ingenieur (NSI)
    +++++++++++++++++++++++++++++++++++++++++++++++++

      Biology & Computer Science Activity : conference (every
      year) Address : NSI - TIRF / INPG 46 avenue Felix
      Viallet 38031 Grenoble Cedex FRANCE

   ------------------------------------------------------------------------

16. A: Other sources of information about NNs?
==============================================

    1. Neuron Digest
    ++++++++++++++++

      Internet Mailing List. From the welcome blurb:
      "Neuron-Digest is a list (in digest form) dealing with all
      aspects of neural networks (and any type of network or
      neuromorphic system)" To subscribe, send email to
      [e-mail]
      comp.ai.neural-net readers also find the messages in that
      newsgroup in the form of digests.

    2. Usenet groups comp.ai.neural-nets (Oha!) and
    +++++++++++++++++++++++++++++++++++++++++++++++
      comp.theory.self-org-sys.
      +++++++++++++++++++++++++

      There is a periodic posting on comp.ai.neural-nets sent by
      [e-mail] (Gregory Aharonian) about Neural
      Network patents.

    3. Central Neural System Electronic Bulletin Board
    ++++++++++++++++++++++++++++++++++++++++++++++++++

      Modem: 409-737-5222; Sysop: Wesley R. Elsberry; 4160
      Pirates' Beach, Galveston, TX 77554;
      [e-mail]. Many MS-DOS PD and
      shareware simulations, source code, benchmarks,
      demonstration packages, information files; some Unix,
      Macintosh, Amiga related files. Also available are files on
      AI, AI Expert listings 1986-1991, fuzzy logic, genetic
      algorithms, artificial life, evolutionary biology, and many
      Project Gutenberg and Wiretap etexts. No user fees have
      ever been charged. Home of the NEURAL_NET Echo,
      available thrugh FidoNet, RBBS-Net, and other EchoMail
      compatible bulletin board systems.

    4. Neural ftp archive site ftp.funet.fi
    +++++++++++++++++++++++++++++++++++++++

      Is administrating a large collection of neural network
      papers and software at the Finnish University Network file
      archive site ftp.funet.fi in directory /pub/sci/neural Contains
      all the public domain software and papers that they have
      been able to find. All of these files have been transferred
      from FTP sites in U.S. and are mirrored about every 3
      months at fastest. Contact: [e-mail]

    5. USENET newsgroup comp.org.issnnet
    ++++++++++++++++++++++++++++++++++++

      Forum for discussion of academic/student-related issues in
      NNs, as well as information on ISSNNet (see answer 12)
      and its activities.

    6. AI CD-ROM
    ++++++++++++

      Network Cybernetics Corporation produces the "AI
      CD-ROM". It is an ISO-9660 format CD-ROM and
      contains a large assortment of software related to artificial
      intelligence, artificial life, virtual reality, and other topics.
      Programs for OS/2, MS-DOS, Macintosh, UNIX, and
      other operating systems are included. Research papers,
      tutorials, and other text files are included in ASCII, RTF,
      and other universal formats. The files have been collected
      from AI bulletin boards, Internet archive sites, University
      computer deptartments, and other government and civilian
      AI research organizations. Network Cybernetics
      Corporation intends to release annual revisions to the AI
      CD-ROM to keep it up to date with current developments
      in the field. The AI CD-ROM includes collections of files
      that address many specific AI/AL topics including Neural
      Networks (Source code and executables for many different
      platforms including Unix, DOS, and Macintosh. ANN
      development tools, example networks, sample data,
      tutorials. A complete collection of Neural Digest is included
      as well.) The AI CD-ROM may be ordered directly by
      check, money order, bank draft, or credit card from:
      Network Cybernetics Corporation; 4201 Wingren Road
      Suite 202; Irving, TX 75062-2763; Tel [phone]; Fax
      214/650-1929; The cost is $129 per disc + shipping ($5/disc
      domestic or $10/disc foreign) (See the comp.ai FAQ for
      further details)

    7. NN events server
    +++++++++++++++++++

      There is a WWW page and an FTP Server for
      Announcements of Conferences, Workshops and Other
      Events on Neural Networks at IDIAP in Switzerland.
      WWW-Server:
      http://www.idiap.ch/html/idiap-networks.html,
      FTP-Server: ftp://ftp.idiap.ch/html/NN-events/,

    8. World Wide Web
    +++++++++++++++++

      In World-Wide-Web (WWW, for example via the xmosaic
      program) you can read neural network information for
      instance by opening one of the following uniform resource
      locators (URLs): http://www.neuronet.ph.kcl.ac.uk
      (NEuroNet, King's College, London),
      http://www.eeb.ele.tue.nl (Eindhoven, Netherlands),
      http://www.emsl.pnl.gov:2080/docs/cie/neural/ (Richland,
      Washington),
      http://www.cosy.sbg.ac.at/~rschwaig/rschwaig/projects.html
      (Salzburg, Austria),
      http://http2.sils.umich.edu/Public/nirg/nirg1.html
      (Michigan). http://rtm.science.unitn.it/ Reactive Memory
      Search (Tabu Search) page (Trento, Italy). Many others
      are available too; WWW is changing all the time.

    9. Neurosciences Internet Resource Guide
    ++++++++++++++++++++++++++++++++++++++++

      This document aims to be a guide to existing, free,
      Internet-accessible resources helpful to neuroscientists of
      all stripes. An ASCII text version (86K) is available in the
      Clearinghouse of Subject-Oriented Internet Resource
      Guides as follows:

      anonymous FTP, Gopher, WWW Hypertext

    10. INTCON mailing list
    +++++++++++++++++++++++

      INTCON (Intelligent Control) is a moderated mailing list
      set up to provide a forum for communication and exchange
      of ideas among researchers in neuro-control, fuzzy logic
      control, reinforcement learning and other related subjects
      grouped under the topic of intelligent control. Send your
      subscribe requests to
      [e-mail]

   ------------------------------------------------------------------------

17. A: Freely available software packages for NN
================================================
   simulation?
   ===========

    1. Rochester Connectionist Simulator
    ++++++++++++++++++++++++++++++++++++

      A quite versatile simulator program for arbitrary types of
      neural nets. Comes with a backprop package and a
      X11/Sunview interface. Available via anonymous FTP
      from cs.rochester.edu in directory pub/packages/simulator
      as the files README (8 KB), rcs_v4.2.tar.Z (2.9 MB),

    2. UCLA-SFINX
    +++++++++++++

      ftp retina.cs.ucla.edu [131.179.16.6]; Login name: sfinxftp;
      Password: joshua; directory: pub; files : README;
      sfinx_v2.0.tar.Z; Email info request :
      [e-mail]

    3. NeurDS
    +++++++++

      simulator for DEC systems supporting VT100 terminal.
      available for anonymous ftp from gatekeeper.dec.com
      [16.1.0.2] in directory: pub/DEC as the file
      NeurDS031.tar.Z (111 Kb)

    4. PlaNet5.7 (formerly known as SunNet)
    +++++++++++++++++++++++++++++++++++++++

      A popular connectionist simulator with versions to run
      under X Windows, and non-graphics terminals created by
      Yoshiro Miyata (Chukyo Univ., Japan). 60-page User's
      Guide in Postscript. Send any questions to
      [e-mail] Available for anonymous ftp
      from ftp.ira.uka.de as /pub/neuron/PlaNet5.7.tar.Z (800 kb)
      or from boulder.colorado.edu [128.138.240.1] as
      /pub/generic-sources/PlaNet5.7.tar.Z

    5. GENESIS
    ++++++++++

      GENESIS 1.4.2 (GEneral NEural SImulation System) is a
      general purpose simulation platform which was developed
      to support the simulation of neural systems ranging from
      complex models of single neurons to simulations of large
      networks made up of more abstract neuronal components.
      Most current GENESIS applications involve realistic
      simulations of biological neural systems. Although the
      software can also model more abstract networks, other
      simulators are more suitable for backpropagation and
      similar connectionist modeling. Available for ftp with the
      following procedure: Use 'telnet' to genesis.bbb.caltech.edu
      and login as the user "genesis" (no password). If you
      answer all the questions, an 'ftp' account will
      automatically be created for you. You can then 'ftp' back to
      the machine and download the software (about 3 MB).
      Contact: [e-mail]. Further information via
      WWW at http://www.bbb.caltech.edu/GENESIS/.

    6. Mactivation
    ++++++++++++++

      A neural network simulator for the Apple Macintosh.
      Available for ftp from ftp.cs.colorado.edu [128.138.243.151]
      as /pub/cs/misc/Mactivation-3.3.sea.hqx

    7. Cascade Correlation Simulator
    ++++++++++++++++++++++++++++++++

      A simulator for Scott Fahlman's Cascade Correlation
      algorithm. Available for ftp from ftp.cs.cmu.edu
      [128.2.206.173] in directory /afs/cs/project/connect/code as
      the file cascor-v1.0.4.shar (218 KB) There is also a version
      of recurrent cascade correlation in the same directory in
      file rcc1.c (108 KB).

    8. Quickprop
    ++++++++++++

      A variation of the back-propagation algorithm developed
      by Scott Fahlman. A simulator is available in the same
      directory as the cascade correlation simulator above in file
      nevprop1.16.shar (137 KB) (see also the description of
      NEVPROP below)

    9. DartNet
    ++++++++++

      DartNet is a Macintosh-based backpropagation simulator,
      developed at Dartmouth by Jamshed Bharucha and Sean
      Nolan as a pedagogical tool. It makes use of the Mac's
      graphical interface, and provides a number of tools for
      building, editing, training, testing and examining networks.
      This program is available by anonymous ftp from
      dartvax.dartmouth.edu [129.170.16.4] as
      /pub/mac/dartnet.sit.hqx (124 KB).

    10. SNNS
    ++++++++

      "Stuttgart Neural Network Simulator" from the University
      of Stuttgart, Germany. A luxurious simulator for many
      types of nets; with X11 interface: Graphical 2D and 3D
      topology editor/visualizer, training visualisation, multiple
      pattern set handling etc. Currently supports
      backpropagation (vanilla, online, with momentum term
      and flat spot elimination, batch, time delay),
      counterpropagation, quickprop, backpercolation 1,
      generalized radial basis functions (RBF), RProp, ART1,
      ART2, ARTMAP, Cascade Correlation, Recurrent
      Cascade Correlation, Dynamic LVQ, Backpropagation
      through time (for recurrent networks), batch
      backpropagation through time (for recurrent networks),
      Quickpropagation through time (for recurrent networks),
      Hopfield networks, Jordan and Elman networks,
      autoassociative memory, self-organizing maps, time-delay
      networks (TDNN), and is user-extendable (user-defined
      activation functions, output functions, site functions,
      learning procedures). Works on SunOS, Solaris, IRIX,
      Ultrix, AIX, HP/UX, and Linux. Available for ftp from
      ftp.informatik.uni-stuttgart.de [129.69.211.2] in directory
      /pub/SNNS as SNNSv3.2.tar.Z (2 MB, Source code) and
      SNNSv3.2.Manual.ps.Z (1.4 MB, Documentation). There
      are also various other files in this directory (e.g. the source
      version of the manual, a Sun Sparc executable, older
      versions of the software, some papers, and the software in
      several smaller parts). It may be best to first have a look at
      the file SNNSv3.2.Readme (10 kb). This file contains a
      somewhat more elaborate short description of the
      simulator.

    11. Aspirin/MIGRAINES
    +++++++++++++++++++++

      Aspirin/MIGRAINES 6.0 consists of a code generator that
      builds neural network simulations by reading a network
      description (written in a language called "Aspirin") and
      generates a C simulation. An interface (called
      "MIGRAINES") is provided to export data from the neural
      network to visualization tools. The system has been ported
      to a large number of platforms. The goal of Aspirin is to
      provide a common extendible front-end language and
      parser for different network paradigms. The MIGRAINES
      interface is a terminal based interface that allows you to
      open Unix pipes to data in the neural network. Users can
      display the data using either public or commercial
      graphics/analysis tools. Example filters are included that
      convert data exported through MIGRAINES to formats
      readable by Gnuplot 3.0, Matlab, Mathematica, and xgobi.
      The software is available from two FTP sites: from CMU's
      simulator collection on pt.cs.cmu.edu [128.2.254.155] in
      /afs/cs/project/connect/code/am6.tar.Z and from UCLA's
      cognitive science machine ftp.cognet.ucla.edu [128.97.50.19]
      in /pub/alexis/am6.tar.Z (2 MB).

    12. Adaptive Logic Network kit
    ++++++++++++++++++++++++++++++

      This package differs from the traditional nets in that it uses
      logic functions rather than floating point; for many tasks,
      ALN's can show many orders of magnitude gain in
      training and performance speed. Anonymous ftp from
      menaik.cs.ualberta.ca [129.128.4.241] in directory
      /pub/atree. See the files README (7 KB), atree2.tar.Z
      (145 kb, Unix source code and examples), atree2.ps.Z (76
      kb, documentation), a27exe.exe (412 kb, MS-Windows 3.x
      executable), atre27.exe (572 kb, MS-Windows 3.x source
      code).

    13. NeuralShell
    +++++++++++++++

      Formerly available from FTP site
      quanta.eng.ohio-state.edu [128.146.35.1] as
      /pub/NeuralShell/NeuralShell.tar". Currently (April 94)
      not available and undergoing a major reconstruction. Not
      to be confused with NeuroShell by Ward System Group
      (see below under commercial software).

    14. PDP
    +++++++

      The PDP simulator package is available via anonymous
      FTP at nic.funet.fi [128.214.6.100] as
      /pub/sci/neural/sims/pdp.tar.Z (202 kb). The simulator is
      also available with the book "Explorations in Parallel
      Distributed Processing: A Handbook of Models, Programs,
      and Exercises" by McClelland and Rumelhart. MIT Press,
      1988. Comment: "This book is often referred to as PDP vol
      III which is a very misleading practice! The book comes
      with software on an IBM disk but includes a makefile for
      compiling on UNIX systems. The version of PDP available
      at ftp.funet.fi seems identical to the one with the book
      except for a bug in bp.c which occurs when you try to run a
      script of PDP commands using the DO command. This can
      be found and fixed easily."

    15. Xerion
    ++++++++++

      Xerion runs on SGI and Sun machines and uses X
      Windows for graphics. The software contains modules that
      implement Back Propagation, Recurrent Back Propagation,
      Boltzmann Machine, Mean Field Theory, Free Energy
      Manipulation, Hard and Soft Competitive Learning, and
      Kohonen Networks. Sample networks built for each of the
      modules are also included. Contact: [e-mail].
      Xerion is available via anonymous ftp from
      ftp.cs.toronto.edu [128.100.1.105] in directory /pub/xerion as
      xerion-3.1.ps.Z (153 kB) and xerion-3.1.tar.Z (1.3 MB)
      plus several concrete simulators built with xerion (about 40
      kB each).

    16. Neocognitron simulator
    ++++++++++++++++++++++++++

      The simulator is written in C and comes with a list of
      references which are necessary to read to understand the
      specifics of the implementation. The unsupervised version
      is coded without (!) C-cell inhibition. Available for
      anonymous ftp from unix.hensa.ac.uk [129.12.21.7] in
      /pub/neocognitron.tar.Z (130 kB).

    17. Multi-Module Neural Computing Environment
    +++++++++++++++++++++++++++++++++++++++++++++
      (MUME)
      ++++++

      MUME is a simulation environment for multi-modules
      neural computing. It provides an object oriented facility for
      the simulation and training of multiple nets with various
      architectures and learning algorithms. MUME includes a
      library of network architectures including feedforward,
      simple recurrent, and continuously running recurrent
      neural networks. Each architecture is supported by a
      variety of learning algorithms. MUME can be used for
      large scale neural network simulations as it provides
      support for learning in multi-net environments. It also
      provide pre- and post-processing facilities. The modules
      are provided in a library. Several "front-ends" or clients
      are also available. X-Window support by
      editor/visualization tool Xmume. MUME can be used to
      include non-neural computing modules (decision trees, ...)
      in applications. MUME is available anonymous ftp on
      mickey.sedal.su.oz.au [129.78.24.170] after signing and
      sending a licence: /pub/license.ps (67 kb). Contact: Marwan
      Jabri, SEDAL, Sydney University Electrical Engineering,
      NSW 2006 Australia, [e-mail]

    18. LVQ_PAK, SOM_PAK
    ++++++++++++++++++++

      These are packages for Learning Vector Quantization and
      Self-Organizing Maps, respectively. They have been built
      by the LVQ/SOM Programming Team of the Helsinki
      University of Technology, Laboratory of Computer and
      Information Science, Rakentajanaukio 2 C, SF-02150
      Espoo, FINLAND There are versions for Unix and
      MS-DOS available from cochlea.hut.fi [130.233.168.48] as
      /pub/lvq_pak/lvq_pak-2.1.tar.Z (340 kB, Unix sources),
      /pub/lvq_pak/lvq_p2r1.exe (310 kB, MS-DOS self-extract
      archive), /pub/som_pak/som_pak-1.2.tar.Z (251 kB, Unix
      sources), /pub/som_pak/som_p1r2.exe (215 kB, MS-DOS
      self-extract archive). (further programs to be used with
      SOM_PAK and LVQ_PAK can be found in /pub/utils).

    19. SESAME
    ++++++++++

      ("Software Environment for the Simulation of Adaptive
      Modular Systems") SESAME is a prototypical software
      implementation which facilitates
       o Object-oriented building blocks approach.
       o Contains a large set of C++ classes useful for neural
         nets, neurocontrol and pattern recognition. No C++
         classes can be used as stand alone, though!
       o C++ classes include CartPole, nondynamic
         two-robot arms, Lunar Lander, Backpropagation,
         Feature Maps, Radial Basis Functions,
         TimeWindows, Fuzzy Set Coding, Potential Fields,
         Pandemonium, and diverse utility building blocks.
       o A kernel which is the framework for the C++
         classes and allows run-time manipulation,
         construction, and integration of arbitrary complex
         and hybrid experiments.
       o Currently no graphic interface for construction, only
         for visualization.
       o Platform is SUN4, XWindows
      Unfortunately no reasonable good introduction has been
      written until now. We hope to have something soon. For
      now we provide papers (eg. NIPS-92), a reference manual
      (>220 pages), source code (ca. 35.000 lines of code), and a
      SUN4-executable by ftp only. Sesame and its description is
      available in various files for anonymous ftp on ftp
      ftp.gmd.de in the directories /gmd/as/sesame and
      /gmd/as/paper. Questions to [e-mail]; there
      is only very limited support available.

    20. Nevada Backpropagation (NevProp)
    ++++++++++++++++++++++++++++++++++++

      NevProp is a free, easy-to-use feedforward
      backpropagation (multilayer perceptron) program. It uses
      an interactive character-based interface, and is distributed
      as C source code that should compile and run on most
      platforms. (Precompiled executables are available for
      Macintosh and DOS.) The original version was Quickprop
      1.0 by Scott Fahlman, as translated from Common Lisp by
      Terry Regier. We added early-stopped training based on a
      held-out subset of data, c index (ROC curve area)
      calculation, the ability to force gradient descent (per-epoch
      or per-pattern), and additional options. FEATURES
      (NevProp version 1.16): UNLIMITED (except by machine
      memory) number of input PATTERNS; UNLIMITED
      number of input, hidden, and output UNITS; Arbitrary
      CONNECTIONS among the various layers' units;
      Clock-time or user-specified RANDOM SEED for initial
      random weights; Choice of regular GRADIENT
      DESCENT or QUICKPROP; Choice of PER-EPOCH or
      PER-PATTERN (stochastic) weight updating;
      GENERALIZATION to a test dataset;
      AUTOMATICALLY STOPPED TRAINING based on
      generalization; RETENTION of best-generalizing weights
      and predictions; Simple but useful GRAPHIC display to
      show smoothness of generalization; SAVING of results to
      a file while working interactively; SAVING of weights file
      and reloading for continued training; PREDICTION-only
      on datasets by applying an existing weights file; In addition
      to RMS error, the concordance, or c index is displayed. The
      c index (area under the ROC curve) shows the correctness
      of the RELATIVE ordering of predictions AMONG the
      cases; ie, it is a measure of discriminative power of the
      model. AVAILABILITY: The most updated version of
      NevProp will be made available by anonymous ftp from the
      University of Nevada, Reno: On ftp.scs.unr.edu
      [134.197.10.130] in the directory
      "pub/goodman/nevpropdir", e.g. README.FIRST (45 kb)
      or nevprop1.16.shar (138 kb). Version 2 (not yet released)
      is intended to have some new features: more flexible file
      formatting (including access to external data files; option to
      prerandomize data order; randomized stochastic gradient
      descent; option to rescale predictor (input) variables);
      linear output units as an alternative to sigmoidal units for
      use with continuous-valued dependent variables (output
      targets); cross-entropy (maximum likelihood) criterion
      function as an alternative to square error for use with
      categorical dependent variables
      (classification/symbolic/nominal targets); and interactive
      interrupt to change settings on-the-fly. Limited support is
      available from Phil Goodman ([e-mail]),
      University of Nevada Center for Biomedical Research.

    21. Fuzzy ARTmap
    ++++++++++++++++

      This is just a small example program. Available for
      anonymous ftp from park.bu.edu [128.176.121.56]
      /pub/fuzzy-artmap.tar.Z (44 kB).

    22. PYGMALION
    +++++++++++++

      This is a prototype that stems from an ESPRIT project. It
      implements back-propagation, self organising map, and
      Hopfield nets. Avaliable for ftp from ftp.funet.fi
      [128.214.248.6] as /pub/sci/neural/sims/pygmalion.tar.Z
      (1534 kb). (Original site is imag.imag.fr:
      archive/pygmalion/pygmalion.tar.Z).

    23. Basis-of-AI-backprop
    ++++++++++++++++++++++++

      Earlier versions have been posted in comp.sources.misc and
      people around the world have used them and liked them.
      This package is free for ordinary users but shareware for
      businesses and government agencies ($200/copy, but then
      for this you get the professional version as well). I do
      support this package via email. Some of the highlights are:
       o in C for UNIX and DOS and DOS binaries
       o gradient descent, delta-bar-delta and quickprop
       o extra fast 16-bit fixed point weight version as well
         as a conventional floating point version
       o recurrent networks
       o numerous sample problems
      Available for ftp from ftp.mcs.com in directory
      /mcsnet.users/drt. Or see the WWW page
      http://www.mcs.com/~drt/home.html. The expanded
      professional version is $30/copy for ordinary individuals
      including academics and $200/copy for businesses and
      government agencies (improved user interface, more
      activation functions, networks can be read into your own
      programs, dynamic node creation, weight decay,
      SuperSAB). More details can be found in the
      documentation for the student version. Contact: Don
      Tveter; 5228 N. Nashville Ave.; Chicago, Illinois 60656;
      [e-mail]

    24. Matrix Backpropagation
    ++++++++++++++++++++++++++

      MBP (Matrix Back Propagation) is a very efficient
      implementation of the back-propagation algorithm for
      current-generation workstations. The algorithm includes a
      per-epoch adaptive technique for gradient descent. All the
      computations are done through matrix multiplications and
      make use of highly optimized C code. The goal is to reach
      almost peak-performances on RISCs with superscalar
      capabilities and fast caches. On some machines (and with
      large networks) a 30-40x speed-up can be measured with
      respect to conventional implementations. The software is
      available by anonymous ftp from risc6000.dibe.unige.it
      [130.251.89.154] as /pub/MBPv1.1.tar.Z (Unix version),
      /pub/MBPv11.zip.Z (MS-DOS version), /pub/mpbv11.ps
      (Documentation). For more information, contact Davide
      Anguita ([e-mail]).

    25. WinNN
    +++++++++

      WinNN is a shareware Neural Networks (NN) package for
      windows 3.1. WinNN incorporates a very user friendly
      interface with a powerful computational engine. WinNN is
      intended to be used as a tool for beginners and more
      advanced neural networks users, it provides an alternative
      to using more expensive and hard to use packages. WinNN
      can implement feed forward multi-layered NN and uses a
      modified fast back-propagation for training. Extensive on
      line help. Has various neuron functions. Allows on the fly
      testing of the network performance and generalization. All
      training parameters can be easily modified while WinNN is
      training. Results can be saved on disk or copied to the
      clipboard. Supports plotting of the outputs and weight
      distribution. Available for ftp from winftp.cica.indiana.edu
      as /pub/pc/win3/programr/winnn093.zip (545 kB).

    26. BIOSIM
    ++++++++++

      BIOSIM is a biologically oriented neural network
      simulator. Public domain, runs on Unix (less powerful
      PC-version is available, too), easy to install, bilingual
      (german and english), has a GUI (Graphical User
      Interface), designed for research and teaching, provides
      online help facilities, offers controlling interfaces, batch
      version is available, a DEMO is provided.
      REQUIREMENTS (Unix version): X11 Rel. 3 and above,
      Motif Rel 1.0 and above, 12 MB of physical memory,
      recommended are 24 MB and more, 20 MB disc space.
      REQUIREMENTS (PC version): PC-compatible with MS
      Windows 3.0 and above, 4 MB of physical memory,
      recommended are 8 MB and more, 1 MB disc space. Four
      neuron models are implemented in BIOSIM: a simple
      model only switching ion channels on and off, the original
      Hodgkin-Huxley model, the SWIM model (a modified HH
      model) and the Golowasch-Buchholz model. Dendrites
      consist of a chain of segments without bifurcation. A
      neural network can be created by using the interactive
      network editor which is part of BIOSIM. Parameters can
      be changed via context sensitive menus and the results of
      the simulation can be visualized in observation windows
      for neurons and synapses. Stochastic processes such as
      noise can be included. In addition, biologically orientied
      learning and forgetting processes are modeled, e.g.
      sensitization, habituation, conditioning, hebbian learning
      and competitive learning. Three synaptic types are
      predefined (an excitatatory synapse type, an inhibitory
      synapse type and an electrical synapse). Additional synaptic
      types can be created interactively as desired. Available for
      ftp from ftp.uni-kl.de in directory /pub/bio/neurobio: Get
      /pub/bio/neurobio/biosim.readme (2 kb) and
      /pub/bio/neurobio/biosim.tar.Z (2.6 MB) for the Unix
      version or /pub/bio/neurobio/biosimpc.readme (2 kb) and
      /pub/bio/neurobio/biosimpc.zip (150 kb) for the PC version.
      Contact: Stefan Bergdoll; Department of Software
      Engineering (ZXA/US); BASF Inc.; D-67056
      Ludwigshafen; Germany; [e-mail]; phone
      0621-60-21372; fax [phone]

    27. The Brain
    +++++++++++++

      The Brain is an advanced neural network simulator for
      PCs that is simple enough to be used by non-technical
      people, yet sophisticated enough for serious research work.
      It is based upon the backpropagation learning algorithm.
      Three sample networks are included. The documentation
      included provides you with an introduction and overview of
      the concepts and applications of neural networks as well as
      outlining the features and capabilities of The Brain. The
      Brain requires 512K memory and MS-DOS or PC-DOS
      version 3.20 or later (versions for other OS's and machines
      are available). A 386 (with maths coprocessor) or higher is
      recommended for serious use of The Brain. Shareware
      payment required. Demo version is restricted to number of
      units the network can handle due to memory contraints on
      PC's. Registered version allows use of extra memory.
      External documentation included: 39Kb, 20 Pages. Source
      included: No (Source comes with registration). Available
      via anonymous ftp from ftp.tu-clausthal.de as
      /pub/msdos/science/brain12.zip (78 kb) and from
      ftp.technion.ac.il as /pub/contrib/dos/brain12.zip (78 kb)
      Contact: David Perkovic; DP Computing; PO Box 712;
      Noarlunga Center SA 5168; Australia; Email:
      [e-mail] (preferred) or [e-mail] or
      [e-mail]

    28. FuNeGen 1.0
    +++++++++++++++

      FuNeGen is a MLP based software program to generate
      fuzzy rule based classifiers. A limited version (maximum of
      7 inputs and 3 membership functions for each input) for
      PCs is available for anonymous ftp from
      obelix.microelectronic.e-technik.th-darmstadt.de in
      directory /pub/neurofuzzy. For further information see the
      file read.me. Contact: Saman K. Halgamuge

    29. NeuDL -- Neural-Network Description Language
    ++++++++++++++++++++++++++++++++++++++++++++++++

      NeuDL is a description language for the design, training,
      and operation of neural networks. It is currently limited to
      the backpropagation neural-network model; however, it
      offers a great deal of flexibility. For example, the user can
      explicitly specify the connections between nodes and can
      create or destroy connections dynamically as training
      progresses. NeuDL is an interpreted language resembling C
      or C++. It also has instructions dealing with
      training/testing set manipulation as well as neural network
      operation. A NeuDL program can be run in interpreted
      mode or it can be automatically translated into C++ which
      can be compiled and then executed. The NeuDL interpreter
      is written in C++ and can be easly extended with new
      instructions. NeuDL is available from the anonymous ftp
      site at The University of Alabama: cs.ua.edu (130.160.44.1)
      in the file /pub/neudl/NeuDLver021.tar. The tarred file
      contains the interpreter source code (in C++) a user
      manual, a paper about NeuDL, and about 25 sample
      NeuDL programs. A document demonstrating NeuDL's
      capabilities is also available from the ftp site:
      /pub/neudl/NeuDL/demo.doc /pub/neudl/demo.doc. For
      more information contact the author: Joey Rogers
      ([e-mail]).

    30. NeoC Explorer (Pattern Maker included)
    ++++++++++++++++++++++++++++++++++++++++++

      The NeoC software is an implementation of Fukushima's
      Neocognitron neural network. Its purpose is to test the
      model and to facilitate interactivity for the experiments.
      Some substantial features: GUI, explorer and tester
      operation modes, recognition statistics, performance
      analysis, elements displaying, easy net construction. PLUS,
      a pattern maker utility for testing ANN: GUI, text file
      output, transformations. Available for anonymous FTP
      from OAK.Oakland.Edu (141.210.10.117) as
      /SimTel/msdos/neurlnet/neocog10.zip (193 kB, DOS
      version)

    31. AINET
    +++++++++

      aiNet is a shareware Neural Networks (NN) application
      for MS-Windows 3.1. It does not require learning, has no
      limits in parameters (input & output neurons), no limits in
      sample size. It is not sensitive toward noise in the data.
      Database can be changed dynamically. It provides a way to
      estimate the rate of error in your prediction. Missing values
      are handled automatically. It has graphical
      spreadsheet-like user interface and on-line help system. It
      provides also several different charts types. aiNet manual
      (90 pages) is divided into: "User's Guide", "Basics About
      Modeling with the AINET", "Examples". Special
      requirements: Windows 3.1, VGA or better. Can be
      downloaded from
      ftp://ftp.cica.indiana.edu/pub/pc/win3/programr/ainet100.zip
      or from
      ftp://oak.oakland.edu/SimTel/win3/math/ainet100.zip

   For some of these simulators there are user mailing lists. Get the
   packages and look into their documentation for further info.

   If you are using a small computer (PC, Mac, etc.) you may want
   to have a look at the Central Neural System Electronic Bulletin
   Board (see answer 13). Modem: 409-737-5312; Sysop: Wesley R.
   Elsberry; 4160 Pirates' Beach, Galveston, TX, USA;
   [e-mail]. There are lots of small simulator
   packages, the CNS ANNSIM file set. There is an ftp mirror site
   for the CNS ANNSIM file set at me.uta.edu [129.107.2.20] in the
   /pub/neural directory. Most ANN offerings are in
   /pub/neural/annsim.

   ------------------------------------------------------------------------

18. A: Commercial software packages for NN
==========================================
   simulation?
   ===========

    1. nn/xnn
    +++++++++

           Name: nn/xnn
        Company: Neureka ANS
        Address: Klaus Hansens vei 31B
                 5037 Solheimsviken
                 NORWAY
          Phone:   [phone]
          Email:   [e-mail]
        Basic capabilities:
         Neural network development tool. nn is a language for specification of
         neural network simulators. Produces C-code and executables for the
         specified models, therefore ideal for application development. xnn is
         a graphical front-end to nn and the simulation code produced by nn.
         Gives graphical representations in a number of formats of any
         variables during simulation run-time. Comes with a number of
         pre-implemented models, including: Backprop (several variants), Self
         Organizing Maps, LVQ1, LVQ2, Radial Basis Function Networks,
         Generalized Regression Neural Networks, Jordan nets, Elman nets,
         Hopfield, etc.
        Operating system: nn: UNIX or MS-DOS, xnn: UNIX/X-windows
        System requirements: 10 Mb HD, 2 Mb RAM
        Approx. price: USD 2000,-

    2. BrainMaker
    +++++++++++++

              Name: BrainMaker, BrainMaker Pro
           Company: California Scientific Software
           Address: 10024 Newtown rd, Nevada City, CA, 95959 USA
         Phone,Fax: [phone]
             Email:  calsci![e-mail] (flakey connection)
        Basic capabilities:  train backprop neural nets
        Operating system:   DOS, Windows, Mac
        System requirements:
        Uses XMS or EMS for large models(PCs only): Pro version
        Approx. price:  $195, $795

        BrainMaker Pro 3.0 (DOS/Windows)     $795
            Gennetic Training add-on         $250
          ainMaker 3.0 (DOS/Windows/Mac)     $195
            Network Toolkit add-on           $150
        BrainMaker 2.5 Student version       (quantity sales only, about $38 each)

        BrainMaker Pro C30 Accelerator Board
                  w/ 5Mb memory              $9750
                  w/32Mb memory              $13,000

        Intel iNNTS NN Development System    $11,800
             Intel EMB Multi-Chip Board      $9750
             Intel 80170 chip set            $940

        Introduction To Neural Networks book $30

        California Scientific Software can be reached at:
        Phone: [phone]     Fax: [phone]    Tech Support: 916 478 9035
        Mail: 10024 newtown rd, Nevada City, CA, 95959, USA
        30 day money back guarantee, and unlimited free technical support.
        BrainMaker package includes:
         The book Introduction to Neural Networks
         BrainMaker Users Guide and reference manual
             300 pages , fully indexed, with tutorials, and sample networks
         Netmaker
             Netmaker makes building and training Neural Networks easy, by
             importing and automatically creating BrainMaker's Neural Network
             files.  Netmaker imports Lotus, Excel, dBase, and ASCII files.
         BrainMaker
             Full menu and dialog box interface, runs Backprop at 750,000 cps
             on a 33Mhz 486.
        ---Features ("P" means is avaliable in professional version only):
        Pull-down Menus, Dialog Boxes, Programmable Output Files,
        Editing in BrainMaker,  Network Progress Display (P),
        Fact Annotation,  supports many printers,  NetPlotter,
        Graphics Built In (P),  Dynamic Data Exchange (P),
        Binary Data Mode, Batch Use Mode (P), EMS and XMS Memory (P),
        Save Network Periodically,  Fastest Algorithms,
        512 Neurons per Layer (P: 32,000), up to 8 layers,
        Specify Parameters by Layer (P), Recurrence Networks (P),
        Prune Connections and Neurons (P),  Add Hidden Neurons In Training,
        Custom Neuron Functions,  Testing While Training,
        Stop training when...-function (P),  Heavy Weights (P),
        Hypersonic Training,  Sensitivity Analysis (P),  Neuron Sensitivity (P),
        Global Network Analysis (P),  Contour Analysis (P),
        Data Correlator (P),  Error Statistics Report,
        Print or Edit Weight Matrices,  Competitor (P), Run Time System (P),
        Chip Support for Intel, American Neurologics, Micro Devices,
        Genetic Training Option (P),  NetMaker,  NetChecker,
        Shuffle,  Data Import from Lotus, dBASE, Excel, ASCII, binary,
        Finacial Data (P),  Data Manipulation,  Cyclic Analysis (P),
        User's Guide quick start booklet,
        Introduction to Neural Networks 324 pp book

    3. SAS Software/ Neural Net add-on
    ++++++++++++++++++++++++++++++++++

             Name: SAS Software
          Company: SAS Institute, Inc.
          Address: SAS Campus Drive, Cary, NC 27513, USA
        Phone,Fax: [phone]
            Email: [e-mail] (Neural net inquiries only)

       Basic capabilities:
         Feedforward nets with numerous training methods
         and loss functions, plus statistical analogs of
         counterpropagation and various unsupervised
         architectures
       Operating system: Lots
       System requirements: Lots
       Uses XMS or EMS for large models(PCs only): Runs under Windows, OS/2
       Approx. price: Free neural net software, but you have to license
                      SAS/Base software and preferably the SAS/OR, SAS/ETS,
                      and/or SAS/STAT products.
       Comments: Oriented toward data analysis and statistical applications

    4. NeuralWorks
    ++++++++++++++

           Name: NeuralWorks Professional II Plus (from NeuralWare)
        Company: NeuralWare Inc.
         Adress: Pittsburgh, PA 15276-9910
          Phone: [phone]
            FAX: [phone]

       Distributor for Europe:
         Scientific Computers GmbH.
         Franzstr. 107, 52064 Aachen
         Germany
         Tel.   (49) [phone]
         Fax.   (49) [phone]
         Email. [e-mail]

       Basic capabilities:
         supports over 30 different nets: backprop, art-1,kohonen,
         modular neural network, General regression, Fuzzy art-map,
         probabilistic nets, self-organizing map, lvq, boltmann,
         bsb, spr, etc...
         Extendable with optional package.
         ExplainNet, Flashcode (compiles net in .c code for runtime),
         user-defined io in c possible. ExplainNet (to eliminate
         extra inputs), pruning, savebest,graph.instruments like
         correlation, hinton diagrams, rms error graphs etc..
       Operating system   : PC,Sun,IBM RS6000,Apple Macintosh,SGI,Dec,HP.
       System requirements: varies. PC:2MB extended memory+6MB Harddisk space.
                            Uses windows compatible memory driver (extended).
                            Uses extended memory.
       Approx. price      : call (depends on platform)
       Comments           : award winning documentation, one of the market
                            leaders in NN software.

    5. MATLAB Neural Network Toolbox (for use with Matlab
    +++++++++++++++++++++++++++++++++++++++++++++++++++++
      4.x)
      ++++

         Contact: The MathWorks, Inc.     Phone: [phone]
                  24 Prime Park Way       FAX: [phone]
                  Natick, MA 01760 email: [e-mail]

      The Neural Network Toolbox is a powerful collection of
      MATLAB functions for the design, training, and
      simulation of neural networks. It supports a wide range of
      network architectures with an unlimited number of
      processing elements and interconnections (up to operating
      system constraints). Supported architectures and training
      methods include: supervised training of feedforward
      networks using the perceptron learning rule, Widrow-Hoff
      rule, several variations on backpropagation (including the
      fast Levenberg-Marquardt algorithm), and radial basis
      networks; supervised training of recurrent Elman
      networks; unsupervised training of associative networks
      including competitive and feature map layers; Kohonen
      networks, self-organizing maps, and learning vector
      quantization. The Neural Network Toolbox contains a
      textbook-quality Users' Guide, uses tutorials, reference
      materials and sample applications with code examples to
      explain the design and use of each network architecture
      and paradigm. The Toolbox is delivered as MATLAB
      M-files, enabling users to see the algorithms and
      implementations, as well as to make changes or create new
      functions to address a specific application.

      (Comment by Richard Andrew Miles Outerbridge,
      [e-mail]:) Matlab is spreading
      like hotcakes (and the educational discounts are very
      impressive). The newest release of Matlab (4.0) ansrwers
      the question "if you could only program in one language
      what would it be?". The neural network toolkit is worth
      getting for the manual alone. Matlab is available with lots
      of other toolkits (signal processing, optimization, etc.) but I
      don't use them much - the main package is more than
      enough. The nice thing about the Matlab approach is that
      you can easily interface the neural network stuff with
      anything else you are doing.

    6. Propagator
    +++++++++++++

        Contact: ARD Corporation,
                 9151 Rumsey Road, Columbia, MD  21045, USA
                 [e-mail]
        Easy to use neural network training package.  A GUI implementation of
        backpropagation networks with five layers (32,000 nodes per layer).
        Features dynamic performance graphs, training with a validation set,
        and C/C++ source code generation.
        For Sun (Solaris 1.x & 2.x, $499),
            PC  (Windows 3.x, $199)
            Mac (System 7.x, $199)
        Floating point coprocessor required, Educational Discount,
        Money Back Guarantee, Muliti User Discount
        Windows Demo on:
          nic.funet.fi        /pub/msdos/windows/demo
          oak.oakland.edu     /pub/msdos/neural_nets
              gatordem.zip    pkzip 2.04g archive file
              gatordem.txt    readme text file

    7. NeuroForecaster
    ++++++++++++++++++

         Name:    NeuroForecaster(TM)/Genetica 3.1
         Contact: Accel Infotech (S) Pte Ltd; 648 Geylang Road;
                  Republic of Singapore 1438; Phone: [phone]; Fax: [phone]
                  [e-mail]
         For IBM PC 386/486 with mouse, or compatibles MS Windows* 3.1,
         MS DOS 5.0 or above 4 MB RAM, 5 MB available harddisk space min;
         3.5 inch floppy drive, VGA monitor or above, Math coprocessor recommended.
           Neuroforecaster 3.1 for Windows is priced at US$1199 per single user
         license.  Please email us ([e-mail]) for order form.
           More information about NeuroForecaster(TM)/Genetical may be found in
         ftp://ftp.technet.sg/Technet/user/accel/nfga40.exe
           NeuroForecaster is a user-friendly neural network program specifically
         designed for building sophisticated and powerful forecasting and
         decision-support systems (Time-Series Forecasting, Cross-Sectional
         Classification, Indicator Analysis)
         Features:
         * GENETICA Net Builder Option for automatic network optimization
         * 12 Neuro-Fuzzy Network Models
         * Multitasking & Background Training Mode
         * Unlimited Network Capacity
         * Rescaled Range Analysis & Hurst Exponent to Unveil Hidden Market
           Cycles & Check for Predictability
         * Correlation Analysis to Compute Correlation Factors to Analyze the
           Significance of Indicators
         * Weight Histogram to Monitor the Progress of Learning
         * Accumulated Error Analysis to Analyze the Strength of Input Indicators
         Its user-friendly interface allows the users to build applications quickly,
         easily and interactively, analyze the data visually and see the results
         immediately.
         The following example applications are included in the package:
         * Credit Rating - for generating the credit rating of bank loan
           applications.
         * Stock market 6 monthly returns forecast
         * Stock selection based on company ratios
         * US$ to Deutschmark exchange rate forecast
         * US$ to Yen exchange rate forecast
         * US$ to SGD exchange rate forecast
         * Property price valuation
         * XOR - a classical problem to show the results are better than others
         * Chaos - Prediction of Mackey-Glass chaotic time series
         * SineWave - For demonstrating the power of Rescaled Range Analysis and
           significance of window size
         Techniques Implemented:
         * GENETICA Net Builder Option - network creation & optimization based on
           Darwinian evolution theory
         * Backprop Neural Networks - the most widely-used training algorithm
         * Fastprop Neural Networks - speeds up training of large problems
         * Radial Basis Function Networks - best for pattern classification problems
         * Neuro-Fuzzy Network
         * Rescaled Range Analysis - computes Hurst exponents to unveil hidden
           cycles & check for predictability
         * Correlation Analysis - to identify significant input indicators

    8. Products of NESTOR, Inc.
    +++++++++++++++++++++++++++

      530 Fifth Avenue; New York, NY 10036; USA; Tel.:
      001-212-398-7955

      Founders: Dr. Leon Cooper (having a Nobel Price) and Dr.
      Charles Elbaum (Brown University). Neural Network
      Models: Adaptive shape and pattern recognition (Restricted
      Coulomb Energy - RCE) developed by NESTOR is one of
      the most powerfull Neural Network Model used in a later
      products. The basis for NESTOR products is the Nestor
      Learning System - NLS. Later are developed: Character
      Learning System - CLS and Image Learning System -
      ILS. Nestor Development System - NDS is a development
      tool in Standard C - one of the most powerfull PC-Tools
      for simulation and development of Neural Networks. NLS
      is a multi-layer, feed forward system with low connectivity
      within each layer and no relaxation procedure used for
      determining an output response. This unique architecture
      allows the NLS to operate in real time without the need for
      special computers or custom hardware. NLS is composed of
      multiple neural networks, each specializing in a subset of
      information about the input patterns. The NLS integrates
      the responses of its several parallel networks to produce a
      system response that is far superior to that of other neural
      networks. Minimized connectivity within each layer results
      in rapid training and efficient memory utilization- ideal for
      current VLSI technology. Intel has made such a chip -
      NE1000.

    9. NeuroShell2/NeuroWindows
    +++++++++++++++++++++++++++

      NeuroShell 2 combines powerful neural network
      architectures, a Windows icon driven user interface, and
      sophisticated utilities for MS-Windows machines. Internal
      format is spreadsheet, and users can specify that
      NeuroShell 2 use their own spreadsheet when editing.
      Includes both Beginner's and Advanced systems, a
      Runtime capability, and a choice of 15 Backpropagation,
      Kohonen, PNN and GRNN architectures. Includes Rules,
      Symbol Translate, Graphics, File Import/Export modules
      (including MetaStock from Equis International) and
      NET-PERFECT to prevent overtraining. Options
      available: Market Technical Indicator Option ($295),
      Market Technical Indicator Option with Optimizer ($590),
      and Race Handicapping Option ($149). NeuroShell price:
      $495.

      NeuroWindows is a programmer's tool in a Dynamic Link
      Library (DLL) that can create as many as 128 interactive
      nets in an application, each with 32 slabs in a single
      network, and 32K neurons in a slab. Includes
      Backpropagation, Kohonen, PNN, and GRNN paradigms.
      NeuroWindows can mix supervised and unsupervised nets.
      The DLL may be called from Visual Basic, Visual C,
      Access Basic, C, Pascal, and VBA/Excel 5. NeuroWindows
      price: $369.

      Contact: Ward Systems Group, Inc.; Executive Park West;
      5 Hillcrest Drive; Frederick, MD 21702; USA; Phone: 301
      662-7950; FAX: [phone]. Contact us for a free demo
      diskette and Consumer's Guide to Neural Networks.

    10. NuTank
    ++++++++++

      NuTank stands for NeuralTank. It is educational and
      entertainment software. In this program one is given the
      shell of a 2 dimentional robotic tank. The tank has various
      I/O devices like wheels, whiskers, optical sensors, smell, fuel
      level, sound and such. These I/O sensors are connected to
      Neurons. The player/designer uses more Neurons to
      interconnect the I/O devices. One can have any level of
      complexity desired (memory limited) and do subsumptive
      designs. More complex design take slightly more fuel, so life
      is not free. All movement costs fuel too. One can also tag
      neuron connections as "adaptable" that adapt their weights
      in acordance with the target neuron. This allows neurons
      to learn. The Neuron editor can handle 3 dimention arrays
      of neurons as single entities with very flexible interconect
      patterns.

      One can then design a scenario with walls, rocks, lights, fat
      (fuel) sources (that can be smelled) and many other such
      things. Robot tanks are then introduced into the Scenario
      and allowed interact or battle it out. The last one alive
      wins, or maybe one just watches the motion of the robots
      for fun. While the scenario is running it can be stopped,
      edited, zoom'd, and can track on any robot.

      The entire program is mouse and graphicly based. It uses
      DOS and VGA and is written in TurboC++. There will
      also be the ability to download designs to another computer
      and source code will be available for the core neural
      simulator. This will allow one to design neural systems and
      download them to real robots. The design tools can handle
      three dimentional networks so will work with video camera
      inputs and such. Eventualy I expect to do a port to UNIX
      and multi thread the sign. I also expect to do a Mac port
      and maybe NT or OS/2

      Copies of NuTank cost $50 each. Contact: Richard Keene;
      Keene Educational Software;
      [e-mail]

      NuTank shareware with the Save options disabled is
      available via anonymous ftp from the Internet, see the file
      /pub/incoming/nutank.readme on the host
      cher.media.mit.edu.

    11. Neuralyst
    +++++++++++++

      Name: Neuralyst Version 1.4; Company: Cheshire
      Engineering Corporation; Address: 650 Sierra Madre Villa,
      Suite 201, Pasedena CA 91107; Phone: [phone];
      Fax: [phone];

      Basic capabilities: training of backpropogation neural nets.
      Operating system: Windows or Macintosh running
      Microsoft Excel Spreadsheet. Neuralyst is an add-in
      package for Excel. Approx. price: $195 for windows or
      Mac. Comments: A simple model that is easy to use.
      Integrates nicely into Microsoft Excel. Allows user to
      create, train, and run backprop ANN models entirely
      within an Excel spreadsheet. Provides macro functions that
      can be called from Excel macro's, allowing you to build a
      custom Window's interface using Excel's macro language
      and Visual Basic tools. The new version 1.4 includes a
      genetic algorithm to guide the training process. A good
      bargain to boot. (Comments by Duane Highley, a user and
      NOT the program developer.
      [e-mail])

    12. NeuFuz4
    +++++++++++

      Name: NeuFuz4 Company: National Semiconductor
      Corporation Address: 2900 Semiconductor Drive, Santa
      Clara, CA, 95052, or: Industriestrasse 10, D-8080
      Fuerstenfeldbruck, Germany, or: Sumitomo Chemical
      Engineering Center, Bldg. 7F 1-7-1, Nakase, Mihama-Ku,
      Chiba-City, Ciba Prefecture 261, JAPAN, or: 15th Floor,
      Straight Block, Ocean Centre, 5 Canton Road, Tsim Sha
      Tsui East, Kowloon, Hong Kong, Phone: [phone]
      (Americas), : 011-49-8141-103-0 Germany :
      0l1-81-3-3299-7001 Japan : [phone] Hong Kong
      Email: [e-mail] (Neural net inquiries only)
      URL:
      http://www.commerce.net/directories/participants/ns/home.html
      Basic capabilities: Uses backpropagation techniques to
      initially select fuzzy rules and membership functions. The
      result is a fuzzy associative memory (FAM) which
      implements an approximation of the training data.
      Operating Systems: 486DX-25 or higher with math
      co-processor DOS 5.0 or higher with Windows 3.1, mouse,
      VGA or better, minimum 4 MB RAM, and parallel port.
      Approx. price : depends on version - see below. Comments
      : Not for the serious Neural Network researcher, but good
      for a person who has little understanding of Neural Nets -
      and wants to keep it that way. The systems are aimed at
      low end controls applications in automotive, industrial, and
      appliance areas. NeuFuz is a neural-fuzzy technology
      which uses backpropagation techniques to initially select
      fuzzy rules and membership functions. Initial stages of
      design using NeuFuz technology are performed using
      training data and backpropagation. The result is a fuzzy
      associative memory (FAM) which implements an
      approximation of the training data. By implementing a
      FAM, rather than a multi-layer perceptron, the designer
      has a solution which can be understood and tuned to a
      particular application using Fuzzy Logic design techniques.
      There are several different versions, some with COP8 Code
      Generator (COP8 is National's family of 8-bit
      microcontrollers) and COP8 in-circuit emulator (debug
      module).

    13. Cortex-Pro
    ++++++++++++++

      Cortex-Pro information is on WWW at:
      http://www.neuronet.ph.kcl.ac.uk/neuronet/software/cortex/www1.html.
      You can download a working demo from there. Contact:
      Michael Reiss (
      http://www.mth.kcl.ac.uk/~mreiss/mick.html) email:
      <[e-mail]>.

    14. PARTEK
    ++++++++++

      PARTEK is a powerful, integrated environment for visual
      and quantitative data analysis and pattern recognition.
      Drawing from a wide variety of disciplines including
      Artificial Neural Networks, Fuzzy Logic, Genetic
      Algorithms, and Statistics, PARTEK integrates data
      analysis and modeling tools into an easy to use "point and
      click" system. The following modules are available from
      PARTEK; functions from different modules are integrated
      with each other whereever possible:
       1. The PARTEK/AVB - The Analytical/Visual Base.
         (TM)

              * Analytical Spreadsheet (TM)
                The Analytical Spreadsheet is a powerful and easy to use data analysis,
                transformations, and visualization tool.  Some features include:
                   - import native format ascii/binary data
                   - recognition and resolution of missing data
                   - complete set of common mathematical & statistical functions
                   - contingency table analysis / correspondence analysis
                   - univariate histogram analysis
                   - extensive set of smoothing and normalization transformations
                   - easily and quickly plot color-coded 1-D curves and histograms,
                     2-D, 3-D, and N-D mapped scatterplots, highlighting selected
                     patterns
                   - Command Line (Tcl) and Graphical Interface

              * Pattern Visualization System (TM)
                The Pattern Visualization System offers the most powerful tools for
                visual analysis of the patterns in your data.  Some features include:
                   - automatically maps N-D data down to 3-D for visualization of
                     *all* of your variables at once
                   - hard copy color Postscript output
                   - a variety of color-coding, highlighting, and labeling options
                     allow you to generate meaningful graphics

              * Data Filters
                Filter out selected rows and/or columns of your data for flexible and
                efficient cross-validation, jackknifing, bootstrapping, feature set
                evaluation, and more.

              * Random # Generators
                Generate random numbers from any of the following parameterized
                distributions:
                   - uniform, normal, exponential, gamma, binomial, poisson

              * Many distance/similarity metrics
                Choose the appropriate distance metric for your data:
                   - euclidean, mahalanobis, minkowski, maximum value, absolute value,
                     shape coefficient, cosine coefficient, pearson correlation,
                     rank correlation, kendall's tau, canberra, and bray-curtis

              * Tcl/Tk command line interface

       2. The PARTEK/DSA - Data Structure Analysis
         Module

              * Principal Components Analysis and Regression
                Also known as Eigenvector Projection or Karhunen-Loeve Expansions,
                PCA removes redundant information from your data.
                   - component analysis, correlate PC's with original variables
                   - choice of covariance, correlation, or product dispersion matrices
                   - choice of eigenvector, y-score, and z-score projections
                   - view SCREE and log-eigenvalue plots

              * Cluster Analysis
                Does the data form groups?  How many?  How compact?  Cluster Analysis
                is the tool to answer these questions.
                   - choose between several distance metrics
                   - optionally weight individual patterns
                   - manually or auto-select the cluster number and initial centers
                   - dump cluster counts, mean, cluster to cluster distances,
                     cluster variances, and cluster labeled data to a matrix viewer or
                     the Analytical Spreadsheet for further analysis
                   - visualize n-dimensional clustering
                   - assess goodness of partion using several internal and external
                     criteria metrics

              * N-Dimensional Histogram Analysis
                Among the most inportant questions a researcher needs to know when
                analyzing patterns is whether or not the patterns can distinguish
                different classes of data.  N-D Histogram Analysis is one tool to
                answer this question.
                   - measures histogram overlap in n-dimensional space
                   - automatically find the best subset of features
                   - rank the overlap of your best feature combinations

              * Non-Linear Mapping
                NLM is an iterative algorithm for visually analyzing the structure of
                n-dimensional data.  NLM produces a non-linear mapping of data which
                preserves interpoint distances of n-dimensional data while reducing
                to a lower dimensionality - thus preserving the structure of the data.
                   - visually analyze structure of n-dimensional data
                   - track progress with error curves
                   - orthogonal, PCA, and random initialization

       3. The PARTEK/CP - Classification and Prediction
         Module

              * Multi-Layer Perceptron
                The most popular among the neural pattern recognition tools is the MLP.
                PARTEK takes the MLP to a new dimension, by allowing the network to
                learn by adapting ALL of its parameters to solve a problem.
                   - adapts output bias, neuron activation steepness, and neuron
                     dynamic range, as well as weights and input biases
                   - auto-scaling at input and output - no need to rescale your data
                   - choose between sigmoid, gaussian, linear, or mixture of neurons
                   - learning rate, momentum can be set independently for each parameter
                   - variety of learning methods and network initializations
                   - view color-coded network, error, etc as network trains, tests, runs

              * Learning Vector Quantization
                Because LVQ is a multiple prototype classifier, it adapts to identify
                multiple sub-groups within classes
                   - LVQ1, LVQ2, and LVQ3 training methods
                   - 3 different functions for adapting learning rate
                   - choose between several distance metrics
                   - fuzzy and crisp classifications
                   - set number of prototypes individually for each class

              * Bayesian Classifier
                Bayes methods are the statistical decision theory approach to
                classification.  This classifier uses statistical properties of your
                data to develop a classification model.

      PARTEK is available on HP, IBM, Silicon Graphics, and
      SUN workstations. For more information, send email to
      "[e-mail]" or call [phone].

    15. NeuroSolutions v2.0
    +++++++++++++++++++++++

      NeuroSolutions is a graphical neural network simulation tool. It is
      the only commercial Windows package to support trajectory learning
      with backpropagation through time.  Because of its object-oriented
      design, NeuroSolutions provides the flexibility needed to construct
      a wide range of learning paradigms and network topologies.  Its GUI
      and extensive probing ability streamline the experimentation process
      by providing real-time analysis of the network during learning.

      Construct any neural network belonging to the additive model,
      including locally and globally recurrent systems.  Use a variety of
      unsupervised learning procedures, such as Hebbian, Sanger's, Oja's,
      Competitive and Kohonen.  Implement RBF, PCA, counterpropagation and
      other hybrid network topologies by seamlessly integrating
      supervised and unsupervised learning.

      During learning, animate changes of internal variables such as
      activations, weights, sensitivities and gradients with a variety of
      probes.  Examples are the oscilloscope, spectrum analyzer, 3D state
      space, scatter, 3D surface, matrix and bitmap.

      NeuroSolutions'  NeuralWizard utility automates the neural network
      design process.  Choose between a wide range of neural models. The
      network parameters are dynamically adjusted based on the user's
      training data.  It is a powerful tool used by both beginners and
      researchers alike.

      NeuroSolutions offers advanced features to meet the integration
      needs of neural network developers.  Once a system is designed and
      simulated using the icon-based development environment,
      NeuroSolutions will generate ANSI-compatible C++ source code to be
      compiled and linked into custom applications.  NeuroSolutions can
      also be customized through user-defined DLL's and OLE support.

      Price: $195 - $1995

      Demo copy available from company or by anonymous ftp:
              ftp://oak.oakland.edu/SimTel/win3/neurlnet/ns2demo1.zip
              ftp://oak.oakland.edu/SimTel/win3/neurlnet/ns2demo2.zip

              NeuroDimension, Inc.
              720 S.W. 2nd Ave., Suite 458
              Gainesville FL, 32601
      Phone:  [phone] or
              [phone]
      FAX:    [phone]
      Email:  [e-mail]

   ------------------------------------------------------------------------

19. A: Neural Network hardware?
===============================

   [who will write some short comment on the most important
   HW-packages and chips?]

   The Number 1 of each volume of the journal "Neural Networks"
   has a list of some dozens of suppliers of Neural Network support:
   Software, Hardware, Support, Programming, Design and Service.

   There is an article by J.A. Hegt (Eindhoven) from 1993 available
   from ftp://ftp.urc.tue.nl/pub/neural/hardware_general.ps.gz

   Here is a short list of companies:

    1. HNC, INC.
    ++++++++++++

         5501 Oberlin Drive
         San Diego
         California 92121
         [phone]
            and a second address at
         7799 Leesburg Pike, Suite 900
         Falls Church, Virginia
         22043
         [phone]
            Note: Australian Dist.: Unitronics
                                    Tel : [phone]
                                    Contact: Martin Keye
         HNC markets:
          'Image Document Entry Processing Terminal' - it recognises
          handwritten documents and converts the info to ASCII.
          'ExploreNet 3000' - a NN demonstrator
          'Anza/DP Plus'- a Neural Net board with 25MFlop or 12.5M peak
          interconnects per second.

    2. SAIC (Sience Application International Corporation)
    ++++++++++++++++++++++++++++++++++++++++++++++++++++++

         10260 Campus Point Drive
         MS 71, San Diego
         CA 92121
         [phone]
         Fax: [phone]

    3. Micro Devices
    ++++++++++++++++

         30 Skyline Drive
         Lake Mary
         FL 32746-6201
         [phone]
         MicroDevices makes   MD1220 - 'Neural Bit Slice'
         Each of the products mentioned sofar have very different usages.
         Although this sounds similar to Intel's product, the
         architectures are not.

    4. Intel Corp
    +++++++++++++

         2250 Mission College Blvd
         Santa Clara, Ca 95052-8125
         Attn ETANN, Mail Stop SC9-40
         [phone]
         Intel is making an experimental chip:
         80170NW - Electrically trainable Analog Neural Network (ETANN)
         It has 64 'neurons' on it - almost fully internally connectted
         and the chip can be put in an hierarchial architecture to do 2 Billion
         interconnects per second.
         Support software by
           California Scientific Software
           10141 Evening Star Dr #6
           Grass Valley, CA 95945-9051
           [phone]
         Their product is called 'BrainMaker'.

    5. NeuralWare, Inc
    ++++++++++++++++++

         Penn Center West
         Bldg IV Suite 227
         Pittsburgh
         PA 15276
         They only sell software/simulator but for many platforms.

    6. Tubb Research Limited
    ++++++++++++++++++++++++

         7a Lavant Street
         Peterfield
         Hampshire
         GU32 2EL
         United Kingdom
         Tel: [phone]

    7. Adaptive Solutions Inc
    +++++++++++++++++++++++++

         1400 NW Compton Drive
         Suite 340
         Beaverton, OR 97006
         U. S. A.
         Tel: [phone];   FAX: [phone]

    8. NeuroDynamX, Inc.
    ++++++++++++++++++++

         4730 Walnut St., Suite 101B
         Boulder, CO 80301
         Voice: [phone]   Fax: [phone]
         Internet: [e-mail]
         NDX sells a number neural network hardware products:
         NDX Neural Accelerators: a line of i860-based accelerator cards for
         the PC that give up to 45 million connections per second for use
         with the DynaMind neural network software.
         iNNTS: Intel's 80170NX (ETANN) Neural Network Training System. NDX's president
         was one of the co-designers of this chip.

    9. IC Tech
    ++++++++++

      NEURO-COMPUTING IC's:
      *  DANN050L (dendro-dendritic artificial neural network)
         + 50 neurons fully connected at the input
         + on-chip digital learning capability
         + 6 billion connections/sec peak speed
         + learns 7 x 7 template in < 50 nsec., recalls in < 400 nsec.
         + low power < 100 milli Watts
         + 64-pin package
      *  NCA717D  (neuro correlator array)
         + analog template matching in < 500 nsec.
         + analog input / digital output pins for real-time computation
         + vision applications in stereo and motion computation
         + 40-pin package
      NEURO COMPUTING BOARD:
      *  ICT1050
         + IBM PC compatible or higher
         + with on-board DANN050L
         + digital interface
         + custom configurations available
      Contact:
      IC Tech (Innovative Computing Technologies, Inc.)
      4138 Luff Court
      Okemos, MI 48864
      [phone]
      [e-mail]

   And here is an incomplete overview over known Neural
   Computers with their newest known reference.

   \subsection*{Digital}
   \subsubsection{Special Computers}

   {\bf AAP-2}
   Takumi Watanabe, Yoshi Sugiyama, Toshio Kondo, and Yoshihiro Kitamura.
   Neural network simulation on a massively parallel cellular array
   processor: AAP-2.
   In International Joint Conference on Neural Networks, 1989.

   {\bf ANNA}
   B.E.Boser, E.Sackinger, J.Bromley, Y.leChun, and L.D.Jackel.\\
   Hardware Requirements for Neural Network Pattern Classifiers.\\
   In {\it IEEE Micro}, 12(1), pages 32-40, February 1992.

   {\bf Analog Neural Computer}
   Paul Mueller et al.
   Design and performance of a prototype analog neural computer.
   In Neurocomputing, 4(6):311-323, 1992.

   {\bf APx -- Array Processor Accelerator}\\
   F.Pazienti.\\
   Neural networks simulation with array processors.
   In {\it Advanced Computer Technology, Reliable Systems and Applications;
   Proceedings of the 5th Annual Computer Conference}, pages 547-551.
   IEEE Comput. Soc. Press, May 1991. ISBN: 0-8186-2141-9.

   {\bf ASP -- Associative String Processor}\\
   A.Krikelis.\\
   A novel massively associative processing architecture for the
   implementation artificial neural networks.\\
   In {\it 1991 International Conference on Acoustics, Speech and
   Signal Processing}, volume 2, pages 1057-1060. IEEE Comput. Soc. Press,
   May 1991.

   {\bf BSP400}
   Jan N.H. Heemskerk, Jacob M.J. Murre, Jaap Hoekstra, Leon H.J.G.
   Kemna, and Patrick T.W. Hudson.
   The bsp400: A modular neurocomputer assembled from 400 low-cost
   microprocessors.
   In International Conference on Artificial Neural Networks. Elsevier
   Science, 1991.

   {\bf BLAST}\\
   J.G.Elias, M.D.Fisher, and C.M.Monemi.\\
   A multiprocessor machine for large-scale neural network simulation.
   In {\it IJCNN91-Seattle: International Joint Conference on Neural
   Networks}, volume 1, pages 469-474. IEEE Comput. Soc. Press, July 1991.
   ISBN: 0-7883-0164-1.

   {\bf CNAPS Neurocomputer}\\
   H.McCartor\\
   Back Propagation Implementation on the Adaptive Solutions CNAPS
   Neurocomputer.\\
   In {\it Advances in Neural Information Processing Systems}, 3, 1991.

   {\bf GENES~IV and MANTRA~I}\\
   Paolo Ienne and  Marc A. Viredaz\\
   {GENES~IV}: A Bit-Serial Processing Element for a Multi-Model
      Neural-Network Accelerator\\
   Proceedings of the International Conference on Application Specific Array
      Processors, Venezia, 1993.

   {\bf MA16 -- Neural Signal Processor}
   U.Ramacher, J.Beichter, and N.Bruls.\\
   Architecture of a general-purpose neural signal processor.\\
   In {\it IJCNN91-Seattle: International Joint Conference on Neural
   Networks}, volume 1, pages 443-446. IEEE Comput. Soc. Press, July 1991.
   ISBN: 0-7083-0164-1.

   {\bf MANTRA I}\\
   Marc A. Viredaz\\
   {MANTRA~I}: An {SIMD} Processor Array for Neural Computation
   Proceedings of the Euro-ARCH'93 Conference, {M\"unchen}, 1993.

   {\bf Mindshape}
   Jan N.H. Heemskerk, Jacob M.J. Murre Arend Melissant, Mirko Pelgrom,
   and Patrick T.W. Hudson.
   Mindshape: a neurocomputer concept based on a fractal architecture.
   In International Conference on Artificial Neural Networks. Elsevier
   Science, 1992.

   {\bf mod 2}
   Michael L. Mumford, David K. Andes, and Lynn R. Kern.
   The mod 2 neurocomputer system design.
   In IEEE Transactions on Neural Networks, 3(3):423-433, 1992.

   {\bf NERV}\\
   R.Hauser, H.Horner, R. Maenner, and M.Makhaniok.\\
   Architectural Considerations for NERV - a General Purpose Neural
   Network Simulation System.\\
   In {\it Workshop on Parallel Processing: Logic, Organization and
   Technology -- WOPPLOT 89}, pages 183-195. Springer Verlag, Mars 1989.
   ISBN: 3-5405-5027-5.

   {\bf NP -- Neural Processor}\\
   D.A.Orrey, D.J.Myers, and J.M.Vincent.\\
   A high performance digital processor for implementing large artificial
   neural networks.\\
   In {\it Proceedings of of the IEEE 1991 Custom Integrated Circuits
   Conference}, pages 16.3/1-4. IEEE Comput. Soc. Press, May 1991.
   ISBN: 0-7883-0015-7.

   {\bf RAP -- Ring Array Processor }\\
   N.Morgan, J.Beck, P.Kohn, J.Bilmes, E.Allman, and J.Beer.\\
   The ring array processor: A multiprocessing peripheral for connectionist
   applications. \\
   In {\it Journal of Parallel and Distributed Computing}, pages
   248-259, April 1992.

   {\bf RENNS -- REconfigurable Neural Networks Server}\\
   O.Landsverk, J.Greipsland, J.A.Mathisen, J.G.Solheim, and L.Utne.\\
   RENNS - a Reconfigurable Computer System for Simulating Artificial
   Neural Network Algorithms.\\
   In {\it Parallel and Distributed Computing Systems, Proceedings of the
   ISMM 5th International Conference}, pages 251-256. The International
   Society for Mini and Microcomputers - ISMM, October 1992.
   ISBN: 1-8808-4302-1.

   {\bf SMART -- Sparse Matrix Adaptive and Recursive Transforms}\\
   P.Bessiere, A.Chams, A.Guerin, J.Herault, C.Jutten, and J.C.Lawson.\\
   From Hardware to Software: Designing a ``Neurostation''.\\
   In {\it VLSI design of Neural Networks}, pages 311-335, June 1990.

   {\bf SNAP -- Scalable Neurocomputer Array Processor}
   E.Wojciechowski.\\
   SNAP: A parallel processor for implementing real time neural networks.\\
   In {\it Proceedings of the IEEE 1991 National Aerospace and Electronics
   Conference; NAECON-91}, volume 2, pages 736-742. IEEE Comput.Soc.Press,
   May 1991.

   {\bf Toroidal Neural Network Processor}\\
   S.Jones, K.Sammut, C.Nielsen, and J.Staunstrup.\\
   Toroidal Neural Network: Architecture and Processor Granularity
   Issues.\\
   In {\it VLSI design of Neural Networks}, pages 229-254, June 1990.

   {\bf SMART and SuperNode}
   P. Bessi`ere, A. Chams, and P. Chol.
   MENTAL : A virtual machine approach to artificial neural networks
   programming. In NERVES, ESPRIT B.R.A. project no 3049, 1991.


   \subsubsection{Standard Computers}

   {\bf EMMA-2}\\
   R.Battiti, L.M.Briano, R.Cecinati, A.M.Colla, and P.Guido.\\
   An application oriented development environment for Neural Net models on
   multiprocessor Emma-2.\\
   In {\it Silicon Architectures for Neural Nets; Proceedings for the IFIP
   WG.10.5 Workshop}, pages 31-43. North Holland, November 1991.
   ISBN: 0-4448-9113-7.

   {\bf iPSC/860 Hypercube}\\
   D.Jackson, and D.Hammerstrom\\
   Distributing Back Propagation Networks Over the Intel iPSC/860
   Hypercube}\\
   In {\it IJCNN91-Seattle: International Joint Conference on Neural
   Networks}, volume 1, pages 569-574. IEEE Comput. Soc. Press, July 1991.
   ISBN: 0-7083-0164-1.

   {\bf SCAP -- Systolic/Cellular Array Processor}\\
   Wei-Ling L., V.K.Prasanna, and K.W.Przytula.\\
   Algorithmic Mapping of Neural Network Models onto Parallel SIMD
   Machines.\\
   In {\it IEEE Transactions on Computers}, 40(12), pages 1390-1401,
   December 1991. ISSN: 0018-9340.

   ------------------------------------------------------------------------

20. A: Databases for experimentation with NNs?
==============================================

    1. The neural-bench Benchmark collection
    ++++++++++++++++++++++++++++++++++++++++

      Accessible via anonymous FTP on ftp.cs.cmu.edu
      [128.2.206.173] in directory /afs/cs/project/connect/bench. In
      case of problems or if you want to donate data, email
      contact is "[e-mail]". The data sets in
      this repository include the 'nettalk' data, 'two spirals',
      protein structure prediction, vowel recognition, sonar signal
      classification, and a few others.

    2. Proben1
    ++++++++++

      Proben1 is a collection of 12 learning problems consisting
      of real data. The datafiles all share a single simple common
      format. Along with the data comes a technical report
      describing a set of rules and conventions for performing
      and reporting benchmark tests and their results. Accessible
      via anonymous FTP on ftp.cs.cmu.edu [128.2.206.173] as
      /afs/cs/project/connect/bench/contrib/prechelt/proben1.tar.gz.
      and also on ftp.ira.uka.de [129.13.10.90] as
      /pub/neuron/proben.tar.gz. The file is about 1.8 MB and
      unpacks into about 20 MB.

    3. UCI machine learning database
    ++++++++++++++++++++++++++++++++

      Accessible via anonymous FTP on ics.uci.edu [128.195.1.1]
      in directory /pub/machine-learning-databases".

    4. NIST special databases of the National Institute Of
    ++++++++++++++++++++++++++++++++++++++++++++++++++++++
      Standards And Technology:
      +++++++++++++++++++++++++

      Several large databases, each delivered on a CD-ROM.
      Here is a quick list.
       o NIST Binary Images of Printed Digits, Alphas, and
         Text
       o NIST Structured Forms Reference Set of Binary
         Images
       o NIST Binary Images of Handwritten Segmented
         Characters
       o NIST 8-bit Gray Scale Images of Fingerprint Image
         Groups
       o NIST Structured Forms Reference Set 2 of Binary
         Images
       o NIST Test Data 1: Binary Images of Hand-Printed
         Segmented Characters
       o NIST Machine-Print Database of Gray Scale and
         Binary Images
       o NIST 8-Bit Gray Scale Images of Mated
         Fingerprint Card Pairs
       o NIST Supplemental Fingerprint Card Data (SFCD)
         for NIST Special Database 9
       o NIST Binary Image Databases of Census Miniforms
         (MFDB)
       o NIST Mated Fingerprint Card Pairs 2 (MFCP 2)
       o NIST Scoring Package Release 1.0
       o NIST FORM-BASED HANDPRINT
         RECOGNITION SYSTEM
      Here are example descriptions of two of these databases:

      NIST special database 2: Structured Forms Reference Set
      -------------------------------------------------------
      (SFRS)
      ------

      The NIST database of structured forms contains 5,590 full
      page images of simulated tax forms completed using
      machine print. THERE IS NO REAL TAX DATA IN
      THIS DATABASE. The structured forms used in this
      database are 12 different forms from the 1988, IRS 1040
      Package X. These include Forms 1040, 2106, 2441, 4562,
      and 6251 together with Schedules A, B, C, D, E, F and SE.
      Eight of these forms contain two pages or form faces
      making a total of 20 form faces represented in the database.
      Each image is stored in bi-level black and white raster
      format. The images in this database appear to be real forms
      prepared by individuals but the images have been
      automatically derived and synthesized using a computer
      and contain no "real" tax data. The entry field values on
      the forms have been automatically generated by a
      computer in order to make the data available without the
      danger of distributing privileged tax information. In
      addition to the images the database includes 5,590 answer
      files, one for each image. Each answer file contains an
      ASCII representation of the data found in the entry fields
      on the corresponding image. Image format documentation
      and example software are also provided. The uncompressed
      database totals approximately 5.9 gigabytes of data.

      NIST special database 3: Binary Images of Handwritten
      -----------------------------------------------------
      Segmented Characters (HWSC)
      ---------------------------

      Contains 313,389 isolated character images segmented
      from the 2,100 full-page images distributed with "NIST
      Special Database 1". 223,125 digits, 44,951 upper-case, and
      45,313 lower-case character images. Each character image
      has been centered in a separate 128 by 128 pixel region,
      error rate of the segmentation and assigned classification is
      less than 0.1%. The uncompressed database totals
      approximately 2.75 gigabytes of image data and includes
      image format documentation and example software.

      The system requirements for all databases are a 5.25"
      CD-ROM drive with software to read ISO-9660 format.
      Contact: Darrin L. Dimmick; [e-mail];
      [phone]

      The prices of the databases are between US$ 250 and 1895
      If you wish to order a database, please contact: Standard
      Reference Data; National Institute of Standards and
      Technology; 221/A323; Gaithersburg, MD 20899; Phone:
      [phone]; FAX: [phone]

      Samples of the data can be found by ftp on
      sequoyah.ncsl.nist.gov in directory /pub/data A more
      complete description of the available databases can be
      obtained from the same host as /pub/databases/catalog.txt

    5. CEDAR CD-ROM 1: Database of Handwritten Cities,
    ++++++++++++++++++++++++++++++++++++++++++++++++++
      States, ZIP Codes, Digits, and Alphabetic Characters
      ++++++++++++++++++++++++++++++++++++++++++++++++++++

      The Center Of Excellence for Document Analysis and
      Recognition (CEDAR) State University of New York at
      Buffalo announces the availability of CEDAR CDROM 1:
      USPS Office of Advanced Technology The database
      contains handwritten words and ZIP Codes in high
      resolution grayscale (300 ppi 8-bit) as well as binary
      handwritten digits and alphabetic characters (300 ppi
      1-bit). This database is intended to encourage research in
      off-line handwriting recognition by providing access to
      handwriting samples digitized from envelopes in a working
      post office.

           Specifications of the database include:
           +    300 ppi 8-bit grayscale handwritten words (cities,
                states, ZIP Codes)
                o    5632 city words
                o    4938 state words
                o    9454 ZIP Codes
           +    300 ppi binary handwritten characters and digits:
                o    27,837 mixed alphas  and  numerics  segmented
                     from address blocks
                o    21,179 digits segmented from ZIP Codes
           +    every image supplied with  a  manually  determined
                truth value
           +    extracted from live mail in a  working  U.S.  Post
                Office
           +    word images in the test  set  supplied  with  dic-
                tionaries  of  postal  words that simulate partial
                recognition of the corresponding ZIP Code.
           +    digit images included in test  set  that  simulate
                automatic ZIP Code segmentation.  Results on these
                data can be projected to overall ZIP Code recogni-
                tion performance.
           +    image format documentation and software included

      System requirements are a 5.25" CD-ROM drive with
      software to read ISO-9660 format. For any further
      information, including how to order the database, please
      contact: Jonathan J. Hull, Associate Director, CEDAR, 226
      Bell Hall State University of New York at Buffalo,
      Buffalo, NY 14260; [e-mail] (email)

    6. AI-CD-ROM (see under answer 13)
    ++++++++++++++++++++++++++++++++++

    7. Time series archive
    ++++++++++++++++++++++

      Various datasets of time series (to be used for prediction
      learning problems) are available for anonymous ftp from
      ftp.santafe.edu [192.12.12.1] in /pub/Time-Series".
      Problems are for example: fluctuations in a far-infrared
      laser; Physiological data of patients with sleep apnea; High
      frequency currency exchange rate data; Intensity of a white
      dwarf star; J.S. Bachs final (unfinished) fugue from "Die
      Kunst der Fuge"

      Some of the datasets were used in a prediction contest and
      are described in detail in the book "Time series prediction:
      Forecasting the future and understanding the past", edited
      by Weigend/Gershenfield, Proceedings Volume XV in the
      Santa Fe Institute Studies in the Sciences of Complexity
      series of Addison Wesley (1994).

   ------------------------------------------------------------------------

   That's all folks.

   Acknowledgements: Thanks to all the people who helped to get the stuff
                     above into the posting. I cannot name them all, because
                     I would make far too many errors then. :->

                     No?  Not good?  You want individual credit?
                     OK, OK. I'll try to name them all. But: no guarantee....

     THANKS FOR HELP TO:
   (in alphabetical order of email adresses, I hope)

    o Steve Ward <[e-mail]>
    o Allen Bonde <[e-mail]>
    o Accel Infotech Spore Pte Ltd <[e-mail]>
    o Ales Krajnc <[e-mail]>
    o Alexander Linden <[e-mail]>
    o Matthew David Aldous <[e-mail]>
    o S.Taimi Ames <[e-mail]>
    o Axel Mulder <[e-mail]>
    o [e-mail]
    o Andy Gillanders <[e-mail]>
    o Davide Anguita <[e-mail]>
    o Avraam Pouliakis <[e-mail]>
    o Kim L. Blackwell <[e-mail]>
    o Mohammad Bahrami <[e-mail]>
    o Paul Bakker <[e-mail]>
    o Stefan Bergdoll <[e-mail]>
    o Jamshed Bharucha <[e-mail]>
    o Yijun Cai <[e-mail]>
    o L. Leon Campbell <[e-mail]>
    o Craig Watson <[e-mail]>
    o Yaron Danon <[e-mail]>
    o David Ewing <[e-mail]>
    o David DeMers <[e-mail]>
    o Denni Rognvaldsson <[e-mail]>
    o Duane Highley <[e-mail]>
    o [e-mail]
    o DJ Meyer <[e-mail]>
    o Donald Tveter <[e-mail]>
    o Athanasios Episcopos
      <[e-mail]>
    o Frank Schnorrenberg <[e-mail]>
    o Gary Lawrence Murphy <[e-mail]>
    o [e-mail]
    o Lee Giles <[e-mail]>
    o Glen Clark <opto![e-mail]>
    o Phil Goodman <[e-mail]>
    o [e-mail]
    o Joerg Heitkoetter
      <[e-mail]>
    o Ralf Hohenstein <[e-mail]>
    o Gamze Erten <[e-mail]>
    o Ed Rosenfeld <[e-mail]>
    o Jean-Denis Muller <[e-mail]>
    o Jeff Harpster <uu0979![e-mail]>
    o Jonathan Kamens <[e-mail]>
    o J.J. Merelo <[e-mail]>
    o Jon Gunnar Solheim <[e-mail]>
    o Josef Nelissen <[e-mail]>
    o Joey Rogers <[e-mail]>
    o Subhash Kak <[e-mail]>
    o Ken Karnofsky <[e-mail]>
    o [e-mail]
    o Luke Koops <[e-mail]>
    o William Mackeown <[e-mail]>
    o Mark Plumbley <[e-mail]>
    o Peter Marvit <[e-mail]>
    o [e-mail]
    o Yoshiro Miyata <[e-mail]>
    o Madhav Moganti <[e-mail]>
    o Jyrki Alakuijala <[e-mail]>
    o Michael Reiss <[e-mail]>
    o [e-mail]
    o Maciek Sitnik <[e-mail]>
    o R. Steven Rainwater <[e-mail]>
    o Paolo Ienne <[e-mail]>
    o Paul Keller <[e-mail]>
    o Michael Plonski <[e-mail]>
    o Lutz Prechelt <[e-mail]> [creator of FAQ]
    o Richard Andrew Miles Outerbridge
      <[e-mail]>
    o Robin L. Getz <[e-mail]>
    o Richard Cornelius <[e-mail]>
    o Rob Cunningham <[e-mail]>
    o [e-mail]
    o Stefan Vogt <[e-mail]>
    o Osamu Saito <[e-mail]>
    o Warren Sarle <[e-mail]>
    o Scott Fahlman <[e-mail]>
    o <[e-mail]>
    o Sheryl Cormicle <[e-mail]>
    o Ted Stockwell <[e-mail]>
    o Serge Waterschoot <[e-mail]>
    o Thomas G. Dietterich <[e-mail]>
    o [e-mail]
    o Ulrich Wendl <[e-mail]>
    o M. Verleysen <[e-mail]>
    o Sherif Hashem <[e-mail]>
    o Matthew P Wiener <[e-mail]>
    o Wesley Elsberry <[e-mail]>

   Bye

     Lutz

   Neural network FAQ / Lutz Prechelt, [e-mail]
--
Lutz Prechelt   (http://wwwipd.ira.uka.de/~prechelt/)   | Whenever you
Institut fuer Programmstrukturen und Datenorganisation  | complicate things,
Universitaet Karlsruhe;  76128 Karlsruhe;  Germany      | they get
(Voice: +49/721/608-4068, FAX: [phone])          | less simple.