This book offers a model for concepts and their dynamics. A basic assumptionis that concepts are composed of specified components, which are representedby large binary patterns whose psychological meaning is governed by the interaction between conceptual modules and other functional modules. A recurrent connectionist model is developed in which some inputs are attracted faster than others by an attractor, where convergence times can beinterpreted as decision latencies. The learning rule proposed is extracted from psychological experiments. The rule has the property that that whena context becomes more familiar, the associations between the concepts of the context spontaneously evolve from loose associations to a more taxonomicorganization.
Author(s): Philip R. Van Loocke (eds.)
Series: Lecture Notes in Computer Science 766 : Lecture Notes in Artificial Intelligence
Edition: 1
Publisher: Springer-Verlag Berlin Heidelberg
Year: 1994
Language: English
Pages: 347
Tags: Artificial Intelligence (incl. Robotics)
Introduction....Pages 1-12
Semantic and pre-semantic representations....Pages 13-30
Demarcation of a module for non-verbal representations....Pages 31-50
Feature packages and the representation of function....Pages 51-67
The internal structure of categories....Pages 69-96
A connectionist model and a proposal for a learning rule....Pages 97-133
Prototypes and more general typicality-effects....Pages 135-164
The influence of contexts on typicalities....Pages 165-175
The basic level of taxonomic organization....Pages 177-200
Conceptual Organization and Its Development....Pages 201-201
The empirical psychology of concept development....Pages 201-209
General discussion....Pages 210-254
