home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!sun-barr!ames!purdue!mentor.cc.purdue.edu!noose.ecn.purdue.edu!hermes.ecn.purdue.edu!tag
- From: tag@hermes.ecn.purdue.edu (Tony A. Gibbens)
- Newsgroups: comp.ai.neural-nets
- Subject: local minima
- Keywords: local minima, backpropagation, error rate
- Message-ID: <1992Jul30.205252.4135@noose.ecn.purdue.edu>
- Date: 30 Jul 92 20:52:52 GMT
- References: <4k2xg_j@rpi.edu>
- Sender: tag@ecn.purdue.edu (Tony A. Gibbens)
- Organization: Purdue University Engineering Computer Network
- Lines: 15
-
-
- I am hoping that somebody or perhaps multiple individuals
- will respond to my request. I need a collection of various
- problems likely (yes likely) to get stuck in local minima when
- trained with backpropagation. If possible, I would like these
- problems to be small or at least manageble. What I mean to say
- is that if possible, I would like the training (and testing) data
- sets not to be overwelmingly complex to generate or over 2000 units
- (items, data units?) each.
-
- Please mail me with the responses. I do not always read this group.
- If you wish, any responses I recieve, I will take the time to post.
-
- thankyou very much
- tag@ecn.purdue.edu (Tony A. Gibbens)
-