Xref: utzoo comp.ai.neural-nets:950 alt.cyb-sys:28 Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!uwm.edu!gem.mps.ohio-state.edu!rpi!leah!bingvaxu!cybsys From: cybsys@bingvaxu.cc.binghamton.edu (CYBSYS-L Moderator) Newsgroups: comp.ai.neural-nets,alt.cyb-sys Subject: Re: Generalization Criteria Message-ID: <2450@bingvaxu.cc.binghamton.edu> Date: 20 Sep 89 23:11:00 GMT References: <509@uvaee.ee.virginia.EDU> Reply-To: cybsys@bingvaxu.cc.binghamton.edu.cc.binghamton.edu (CYBSYS-L Moderator) Organization: SUNY Binghamton, NY Lines: 32 Really-From: "Michael H. Prager" This message was originally submitted by MHP100F@ODUVM to the CYBSYS-L list at BINGVMB. On Tue, 19 Sep 89 20:17:26 EDT CYBSYS-L Moderator said: >From: aam9n@uvaee.ee.virginia.EDU (Ali Minai) >[ The following is cross-posted from alt.cyb-sys ] > >Given a *finite* set of input-output values, an estimator/approximator >is used to induce a "reasonable" model for the system that generated >the data. While there are many measures of "goodness of fit" with >respect to the given data set (e.g. mean squared error), there seems >to be no universally accepted measures for the "generalization" achieved.... I believe that this is indeed a very difficult and profound question. One reference of interest is Barron, A. 1984. Predicted squared error: a criterion for automatic model selection. Chapter 4 IN Farlow, S. J., editor. Self-organizing methods in modeling. Marcel Dekker, NY. Barron has also done some work on criteria based on minimum descriptor length, but I am not very familiar with that work. It would be interesting to hear what else you come up with. +--------------------------------------------------------------------+ | Michael H. Prager, Asst. Prof. || FAX: (804) 683-5303 | Department of Oceanography || Bitnet: MHP100F AT ODUVM | Old Dominion University || Internet: mhp100f@oduvm.cc.odu.edu | Norfolk, VA 23529-0276 || Telephone: (804) 683-6003 +--------------------------------------------------------------------+