home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.ai.neural-nets
- Path: sparky!uunet!news.univie.ac.at!blekul11!frmop11!barilvm!tavor!aristo.tau.ac.il!gemini.math.tau.ac.il!moshes
- From: moshes@gemini.math.tau.ac.il (Sipper Moshe)
- Subject: Backpropagation simulation - possible bugs to look for
- Message-ID: <1992Aug31.115035.28451@aristo.tau.ac.il>
- Sender: news@aristo.tau.ac.il
- Nntp-Posting-Host: gemini.math.tau.ac.il
- Organization: School of Math & CS - Tel Aviv University , Tel Aviv , ISRAEL.
- Date: Mon, 31 Aug 1992 11:50:35 GMT
- Lines: 10
-
- I have written a simple backprop simulation
- program in which the weights converge, but alas
- the error rate does not....
-
- Are there maybe some small but important pointers to watch out for ?
- (weight resolution, initialization, logistic function etc' ?)
-
- thanks in advance for any reply.
-
- Moshe Sipper.
-