home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.ai.neural-nets
- Path: sparky!uunet!das.wang.com!wang!news
- From: alex@maxc.technion.ac.il
- Subject: Schnorrenberg Neocognitron Simulator ?
- Organization:
- Date: Thu, 27 Aug 1992 12:10:25 GMT
- Message-ID: <1992Aug27.141025.1@maxc.technion.ac.il>
- Sender: news@wang.com
- Lines: 29
-
- Hello,
- Is anybody experienced with Neocognitron Simulator of F.Schnorrenberg? If
- yes, please could you answer a few questions:
-
- 1. How to obtain during the training process the necessary set of weights,
- matching for classification the given set of patterns. What is the syntax
- of this command?
-
- 2. How to evoke the simulator ONLY to classify (based on the given set of
- weights) the given set of patterns, WITHOUT ANY UPDATING of the weights
- (i.e.training)?
-
- 3. How to explain the screen output of the Simulator, specifically, what
- does it mean that he finds the winners in every stage in the beginning of
- the training process. Known, that self - organization of the Neocognitron
- progresses step by step from the front layer to the last and no need to
- propagate the input pattern from the beginning to the end before training
- is completed. What do the huge arrays that the Simulator prints out mean?
- The weights?
-
- Does anybody know another public domain simulator of the Neocognitron
- (under MS-DOS,Windows, X11, ets) more friendly to use? All the advises,
- references etc will be thankfully appreciated. Please respond directly or
- into the net-news group ai.neural. The responses will be summarized and
- posted. Thanks in advance,
-
- Alex Beylin
- Dept. of Bio-Med. Eng., Technion,
- Haifa, Israel
-