Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!rutgers!ames!ucbcad!ucbvax!atc.bendix.COM!MOSKOWITZ%TSD
From: MOSKOWITZ%TSD@atc.bendix.COM (LEN MOSKOWITZ)
Newsgroups: comp.ai.digest
Subject: Request for assistance
Message-ID: <8705221843.AA24853@ucbvax.Berkeley.EDU>
Date: Thu, 21-May-87 11:43:00 EDT
Article-I.D.: ucbvax.8705221843.AA24853
Posted: Thu May 21 11:43:00 1987
Date-Received: Sat, 23-May-87 16:34:35 EDT
Sender: daemon@ucbvax.BERKELEY.EDU
Distribution: world
Organization: The ARPA Internet
Lines: 21
Approved: ailist@stripe.sri.com


     I'm working on a memory model that learns concepts from scratch.  Given
events consisting of sensory input (e.g. for the vision modality, some
description of scenes), it will (hopefully) learn appropriate groupings of
features that define concepts.  I am looking for sets of primitives that can
describe sensory perceptions.  The primitives need not be "correct" nor
"exhaustive" when evaluated for psychological/perceptual validity, but they
should be "adequate" to describe the range of features they apply to.  I have
one set of visual primitives (Irving Biederman's from SUNY Buffalo's Psych
department) that may handle volumetric descriptions of objects describable by
count nouns.  To fill out the vision primitives, I think I need textural,
motion, size, orientation, and color/brightness/contrast primitives too.  I'm
also looking for perceptual primitives for the other sensory modalities (aural,
tactile, olfactory, kinesthetic...).  Any pointers would be greatly
appreciated.

Len Moskowitz
moskowitz@bendix.com (CSnet)
moskowitz%bendix.com@relay.cs.net (ARPAnet)
moskowit@topaz.rutgers.edu (alternate ARPAnet)
rutgers!topaz!moskowit (uucp)