home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
OS/2 Shareware BBS: 16 Announce
/
16-Announce.zip
/
294250.zip
/
294250.ANN
Wrap
Text File
|
1994-05-03
|
17KB
|
266 lines
DATE 940503
CATEGORY GD00, GD60, G120, GD70, WS00, WS20, WS25, WS50
TITLE IBM NEURAL NETWORK UTILITY ENTRY FOR OS/2 AND WINDOWS
ABSTRACT Neural networks and fuzzy logic are powerful new application
development technologies. Neural networks can find patterns and
learn relationships in business data. They differ from traditional
programming in that the neural network training process is automated,
providing a cost-effective way of turning data into useful processing
functions. Fuzzy logic is an exciting extension of the well-known,
rule-based expert system methodology, with the advantage of letting
experts represent their knowledge in their own language. The Neural
Network Utility (NNU) product family brings the power and
productivity of these advanced technologies to application developers
working in OS/2 (R) and Windows (1) computing environments.
Neural Network Utility Entry for OS/2 (NNU Entry/2) and Neural
Network Utility Entry for Windows (NNU Entry/Win) Version 3.1 are
new, competitively priced, full-function members of the Neural
Network Utility product family. The NNU Entry/2 and NNU Entry/Win
products are functionally equivalent. They can be used as
stand-alone neural network and fuzzy rule system development tools
for educational use or for personal productivity applications.
NNU provides a language-independent API for embedding neural
network and fuzzy rule processing into OS/2 and Windows application
programs.
Refer to the following Software Announcements, dated
May 3, 1994, for additional information.
o 294-248 Neural Network Utility/400
o 294-249 Neural Network Utility for OS/2 and Neural Network
Utility for Windows,
o 294-251 AIX Neural Network Utility/6000
One-Time Charge: $250
Planned Availability Date: September 9, 1994
(R) Registered trademark of International Business Machines
Corporation
(1) Trademark of Microsoft Corporation
IN BRIEF . . .
Features
o Neural network application development support
o Knowledge-based processing using fuzzy rules
o Enhanced application developer productivity tools
Neural Network Utility Family
o Neural Network Utility Entry for OS/2 (NNU Entry/2)
o Neural Network Utility Entry for Windows (NNU Entry/Win)
o Neural Network Utility for OS/2 (NNU/2)
o Neural Network Utility for Windows (NNU/Win)
o Neural Network Utility/400 (NNU/400)
o AIX (R) Neural Network Utility/6000 (NNU/6000)
PRODNO 5871-AAA 03H3894
5871-AAA 03H3895
OVERVIEW DESCRIPTION
NEURAL NETWORK APPLICATION DEVELOPMENT SUPPORT
NNU Entry for OS/2 and NNU Entry for Windows allow users to develop
neural network applications on PCs and deploy them as stand-alone PC
applications.
ENABLE CUSTOMER'S COMPETITIVE ADVANTAGE: Neural network technology
detects patterns and relationships in data and then uses that
information to perform data processing functions. In cases where the
relationships are complex or unknown, neural networks open new areas
to application development. Leading-edge applications using this
technology have provided significant technical advantages in the
retail, marketing, finance, manufacturing, and health industries.
APPLICATION DEVELOPER PRODUCTIVITY: Application development time for
neural network applications can be significantly less than for
traditional data processing applications. A typical neural network
application is developed by doing problem analysis, data analysis and
scaling, training the neural network, and then testing and
validation. There are no algorithm design or code and debug steps in
the neural network application development process, resulting in
shorter development times.
KNOWLEDGE-BASED PROCESSING USING FUZZY RULES
Fuzzy rule processing adds a knowledge-based processing approach to
complement the data-based neural network approach.
ENABLE CUSTOMER'S COMPETITIVE ADVANTAGE: Rule-based expert systems
have been widely used in business applications. The two areas of
weakness are the difficulty maintaining large rule bases and the
sometimes slow execution times. Fuzzy rule systems typically allow a
problem to be solved using fewer rules, and the execution times are
fast because no search algorithms are used. Fuzzy rules are
translated into mathematical form and solved by the fuzzy inference
engine. This technology allows domain experts to represent their
knowledge using their own terminology and quickly develop
knowledge-based applications.
ENABLING NEW APPLICATIONS: The addition of fuzzy rule processing to
the existing neural network capabilities of NNU creates the
possibility of new applications using the pattern recognition and
modeling capabilities of neural networks along with the rule
processing of fuzzy rule systems. The fuzzy rule editor in NNU
allows a developer to quickly define fuzzy variables, fuzzy sets and
membership functions, and corresponding fuzzy rules. This
easy-to-use environment leads to rapid prototyping and evaluation of
the expert-system application.
ENHANCED APPLICATION DEVELOPER PRODUCTIVITY TOOLS: The NNU built-in
data translation tools and scripting language for automated training
have been enhanced to provide improved ease-of-use and increased
function.
APPLICATION DEVELOPER PRODUCTIVITY: The NNU translate editor and
translate filter provide a powerful template-based function for
performing data translation and scaling of neural network training
data. A large percentage of the neural network application
development cycle is spent working with data. NNU splits source
files into training and testing data automatically, generates
translation templates by scanning the source data, provides more than
20 mathematical functions including vector operations, and provides
an easy-to-use editor for specifying the translation operations.
This allows developers to quickly and easily create what-if scenarios
to explore data representation issues.
The NNU control script editor and language allow a developer
to specify an automated sequence of actions for training and testing
neural network applications. Neural network parameters can be set,
data sources can be selected, network performance can be logged, and
breakpoints can be set. New features include text script files that
can be edited in any text editor, variable support so that re-usable
scripts can be written, and looping control constructs. The NNU
control script language allows automated development of neural
network applications.
PRODUCT POSITIONING
The NNU family of products is a complete set of development tools
that combine neural network and fuzzy rule system technologies that
can be used to create leading-edge applications.
The Neural Network Utility Entry products for OS/2 and Windows
are competitively priced, full-function neural network and fuzzy rule
system application development tools. They are intended for
personal, educational, or productivity applications. The NNU Entry
products are not intended for commercial application development.
OS/2 and Windows development environments use graphical user
interfaces to allow point-and-click construction of applications.
NNU Entry products provide tools for data translation and scaling,
visualization of neural networks and fuzzy sets, and a scripting
language for automated training sequences. It supports multiple
neural network models, and provides programming interfaces for
user-defined custom models.
Neural Network Utility can be used across multiple industries
and enables the development of a wide range of applications. NNU
provides a cost-effective way of turning data into useful processing
functions using neural networks and turning knowledge into
application logic using fuzzy rule systems. Applications range from
risk assessment and portfolio management in finance to fraud
detection and policy assessments in health and insurance, all the way
to optimization, scheduling, and visual inspection systems in
manufacturing and process control.
The Neural Network Utility Entry products should be used by
students or application developers who want to explore neural network
and fuzzy rule system capabilities. NNU Entry products can be used
to prototype applications or to develop personal productivity
applications using these technologies.
CUSTINFO PUBLICATIONS
The NEURAL NETWORK UTILITY INSTALLATION AND USER'S GUIDE
(SC41-0223) publication is shipped with the product.
Additional copies of unlicensed publications will be available
for a fee after product availability.
Copies may be ordered from your IBM representative, through
the System Library Subscription Service (SLSS), or by direct order.
ORDER
TITLE NUMBER
Neural Network Utility
Programmers' Reference SC41-0222
EDUCATION SUPPORT
The appropriate curriculum will be updated, as necessary, to
include the enhancements in this announcement.
Details of the education support will be provided by product
availability through separate Education Announcements.
Descriptions of all classroom and self-study courses are
contained in the SKILL DYNAMICS (TM) CATALOG OF IBM EDUCATION
(G320-1244).
Call Skill Dynamics, An IBM Company at 800-IBM-TEACH
(426-8322) for schedules and enrollments.
(TM) Trademark of International Business Machines Corporation
TECHINFO TECHNICAL INFORMATION
SPECIFIED OPERATING ENVIRONMENT
HARDWARE REQUIREMENTS:
NNU ENTRY FOR OS/2: Any personal computer supported by IBM OS/2 2.1,
or later.
o 386DX 33MHz processor (486DX 33MHz recommended)
o 6MB minimum RAM memory (8MB recommended)
o VGA, or higher, resolution display supported by OS/2
o 4MB hard file available free space
NNU ENTRY FOR WINDOWS: Any personal computer supported by Microsoft
Windows 3.1, or later, compatible version in enhanced mode.
o 386DX 33MHz processor (486DX 33MHz recommended)
o 4MB minimum RAM memory (6MB recommended)
o VGA, or higher, resolution display supported by Windows 3.1
o 4MB hard file available free space
SOFTWARE REQUIREMENTS:
NNU ENTRY FOR OS/2:
o Neural Network Utility Entry for OS/2 Version 3.1 will run only
under OS/2 2.1, or higher.
o Custom neural network model and custom interface program
development requires an OS/2 compiler capable of generating
32-bit OS/2 2.1 compatible DLLs.
o NNU Entry/2 local database support requires a DB2/2 (TM)
database.
o NNU Entry/2 remote database access requires the facilities
provided by the DB2 (R) family of client/server products.
NNU ENTRY FOR WINDOWS:
o Neural Network Utility Entry for Windows Version 3.1 will run
under Microsoft** Windows 3.1, or later, compatible version.
o Custom neural network model and custom interface program
development requires a compiler capable of generating 32-bit
Windows compatible DLLs.
o NNU Entry/Win provides generic (local and remote) database
support through the Open DataBase Connectivity (ODBC) interface.
This requires an ODBC driver for each target database.
COMPATIBILITY: Neural Network Utility Entry for OS/2 and Neural
Network Utility Entry for Windows are compatible with other members
of the NNU 3.1 product family. Neural networks and fuzzy rule
systems developed using NNU Entry/2 or NNU Entry/Win can be run by
other NNU family products.
PERFORMANCE CONSIDERATIONS: Neural network application development
typically requires large amounts of floating-point computations.
Application developer productivity will be gated by the speed of the
hardware. Fuzzy rule system development places no great demands on
the system.
In general, neural network and fuzzy logic applications can be
deployed on any machine capable of running Windows 3.1 or OS/2 2.1.
PLANNING INFORMATION
INSTALLABILITY: Installing NNU Entry/2 and NNU Entry/Win is easy and
can be done in less than 30 minutes. An installation program will
guide the user through the installation process.
PACKAGING: The IBM Neural Network Utility Entry for OS/2 and Neural
Network Utility Entry for Windows contain the following packaged as a
single unit:
PROGRAM PRODUCT:
o Registration Card
o Installation and User Guide (publication)
o IBM Program License Agreement on diskette container
o Program diskettes (3.5-inch media)
SECURITY, AUDITABILITY AND CONTROL
User management is responsible for evaluation, selection and
implementation of security features, administrative procedures and
appropriate controls in application systems and communication
facilities.
ORDERING ORDERING INFORMATION
To order the programs in this announcement for order type,
specify the order type number, feature number, part number, program
name, media, and quantity.
ORDER
TYPE FEATURE PART
NUMBER NUMBER NUMBER PROGRAM NAME MEDIA
5871-AAA 7896 03H3894 NNU Entry
for OS/2 3.5-inch
5871-AAA 7897 03H3895 NNU Entry
for Windows 3.5-inch
o Program Package -- Program diskette with documentation.
TERMS AND CONDITIONS
LICENSING: The programs in this announcement are licensed under the
terms and conditions of the IBM Program License Agreement. A copy of
the agreement is included in the program package.
LIMITED WARRANTY: Program: No Media: Three months
PROGRAM SERVICES: None
VOLUME ORDERS: Not applicable
NATIONAL EDUCATION LICENSE FEE: A National Education License Fee
(NELF) is available under the terms and conditions of the Education
Allowance Attachment. The NELF, which represents a 20% discount off
the single unit price, is not subject to any further discount or
allowance.
CHARGES
PROGRAM FEATURE ONE-TIME
NUMBER NUMBER ORDER TYPE CHARGE
03H3894 7896 NNU Entry for OS/2
Program Package $250
03H3895 7897 NNU Entry for Windows
Program Package 250
The charges are unchanged by this announcement.
- - - E N D O F P R I N T O U T - - -