home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.ai.neural-nets
- Path: sparky!uunet!cs.utexas.edu!sdd.hp.com!zaphod.mps.ohio-state.edu!magnus.acs.ohio-state.edu!usenet.ins.cwru.edu!news.ysu.edu!do-not-reply-to-path
- From: FM010201@ysub.ysu.edu
- Subject: looking for somebody's experience with hect-nelson NN board
- Message-ID: <16842117AA.FM010201@ysub.ysu.edu>
- Sender: news@news.ysu.edu (Usenet News Admin)
- Nntp-Posting-Host: ysub
- Organization: Youngstown State University
- Date: Thu, 13 Aug 1992 23:53:14 GMT
- Lines: 11
-
- I am training a neural net on a sun workstation by using the back propagation a
- lgorithm. The training data set has 31 inputs and 37 outputs with 50 as hidden
- nodes in the hidden layer. The training data set consist of 500 pairs and it ha
- s taken about two weeks of time to reach to a desirable accuracy. I am interest
- ed in finding out about somebody's experience with the accelerator boards which
- improve the performance of back propgation for speed? what is the general spee
- d up in the performance. I also tried radial based function nets. All neu
- ral net formulations seems to work very well on small demo problems ,however, o
- n large realistic problems training time is exorbitant? any thoughts and sugges
- on this. Thanks in advance.
- Alam INTERNET fm010201@ysub.ysu.edu
-