home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!darwin.sura.net!spool.mu.edu!sol.ctr.columbia.edu!eff!ssd.intel.com!ogicse!das-news.harvard.edu!cantaloupe.srv.cs.cmu.edu!crabapple.srv.cs.cmu.edu!news
- From: sef@sef-pmax.slisp.cs.cmu.edu
- Newsgroups: comp.ai.neural-nets
- Subject: Re: CasCor2
- Message-ID: <BzzMu3.F49.1@cs.cmu.edu>
- Date: 28 Dec 92 21:10:50 GMT
- Article-I.D.: cs.BzzMu3.F49.1
- Sender: news@cs.cmu.edu (Usenet News System)
- Organization: School of Computer Science, Carnegie Mellon
- Lines: 41
- Nntp-Posting-Host: sef-pmax.slisp.cs.cmu.edu
-
-
- From: simon@socrates.et.tudelft.nl (Natalio Simon)
-
- Is this new version (CasCor2) of the Cascade-Correlation Algorithm of Scott
- Fahlman available?
-
- Maybe it is posible for Scott Fahlman to point out how CasCor2 differs from
- CasCor1.
-
- I've had several inquiries about "CasCor 2" since it was mentioned in Scott
- Markel's workshop summary, so perhaps it is best to post the information
- here.
-
- The new algorithm is called "Cascade 2", not "Cascade Correlation 2" or
- "Cascor 2" or "<anything> II". (I hate Roman numerals.) The reason for
- the name change is that there's no "correlation" part left. We have known
- for some time that the correlation (really a covariance) measure used in
- Cascor tends to produce hidden units that overshoot small errors. This
- isn't a problem for tasks with binary outputs, but leads to excess units
- and poor generalization on tasks that require continuous-valued analog
- outputs. Cascade 2 uses a different algorithm for training the candidate
- units, basically minimizing the difference between the candidate unit's
- output and the residual error. But there are a few tricky details I don't
- want to explain right now.
-
- I plan to release this C2 code generally as soon as I finish a few more
- tests and write up a tech report on it. I've been too busy scrounging for
- funding for most of the past year to finish that task, but I hope to return
- to it in the near future. As soon as the TR and the code are generally
- available, I'll announce that fact in this newsgroup.
-
- -- Scott
-
- ===========================================================================
- Scott E. Fahlman Internet: sef+@cs.cmu.edu
- Senior Research Scientist Phone: 412 268-2575
- School of Computer Science Fax: 412 681-5739
- Carnegie Mellon University Latitude: 40:26:33 N
- 5000 Forbes Avenue Longitude: 79:56:48 W
- Pittsburgh, PA 15213
- ===========================================================================
-