Accession Number : ADA336777
Title : Mathematical Methods for the Implementation of Neural Networks.
Descriptive Note : Final technical rept. 15 Aug 92-30 Jun 96,
Corporate Author : YALE UNIV NEW HAVEN CT DEPT OF COMPUTER SCIENCE
Personal Author(s) : Mjolsness, Eric
PDF Url : ADA336777
Report Date : 18 DEC 1996
Pagination or Media Count : 8
Abstract : We present a novel optimizing network architecture with applications in vision, learning, pattern recognition and combinatorial optimization. This architecture is constructed by combining the following techniques: (1) deterministic annealing, (2) self-amplification, (3) algebraic transformations, (4) clocked objectives, and (5) soft assign. Deterministic annealing in conjunction with self-amplification avoids poor local minima and ensures that a vertex of the hypercube is reached. Algebraic transformations and clocked objectives help partition the relaxation into distinct phases. The problems considered have doubly stochastic matrix constraints or minor variations thereof. We introduce a new technique, soft assign, which is used to satisfy this constraint. Experimental results on different problems are presented and discussed.
Descriptors : *NEURAL NETS, *NUMERICAL METHODS AND PROCEDURES, *APPLIED MATHEMATICS, ANNEALING, OPTIMIZATION, NETWORKS, ALGEBRA, PATTERN RECOGNITION, VISION, COMBINATORIAL ANALYSIS, LEARNING, DETERMINANTS(MATHEMATICS), TRANSFORMATIONS.
Subject Categories : Numerical Mathematics
Distribution Statement : APPROVED FOR PUBLIC RELEASE