home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!elroy.jpl.nasa.gov!news.claremont.edu!ucivax!moe!kofax!lawnet!pride386!dpdvax!pride386!odetics!gordius!news.service.uci.edu!usc!sdd.hp.com!spool.mu.edu!tulane!rangan
- Newsgroups: comp.ai.neural-nets
- Subject: Green's Function and Neural Network Learning
- Message-ID: <545.2b3a961a@dpd.com>
- From: rangan@cs.tulane.edu (Sudha Rangan)
- Date: 25 Dec 92 05:03:22 PT
- Lines: 25
-
- Nntp-Posting-Host-[nntpd-3754]: hermes
- Message-ID: <1992Dec23.171308.3845@cs.tulane.edu>
- Sender: news@cs.tulane.edu
- Organization: Computer Science Dept., Tulane Univ., New Orleans, LA
- Date: Wed, 23 Dec 1992 17:13:08 GMT
- Lines: 18
-
-
-
- I have a question in regularization theory, which provides
- a framework for interpolation, in neural network learning.
-
- The function that minimizes the error function
- is a linear combination of several Green's functions.
- The coefficients of this linear combination satisfy a system
- of linear equations, where the coefficient matrix depends on
- evaluations of Green's functions at specific points.
-
- The problem is Green's function can be singular in certain
- circumstances. Does it mean regularization approach works
- only for the case when Green's functions are non-singular,
- or there are other approaches?
-
- Thanks in advance and Merry X'mas and Happy New Year!
-
-