Information storage capacity of incompletely connected associative memories

被引:20
作者
Bosch, H
Kurfess, FJ [1 ]
机构
[1] New Jersey Inst Technol, Dept Comp & Informat Sci, Newark, NJ 07102 USA
[2] Univ Geneva, Dept Comp Sci, Geneva, Switzerland
关键词
associative memory; information storage capacity; incomplete interconnectivity; hetero-association; auto-association; hebbian learning; retrieval strategy; sparse coding;
D O I
10.1016/S0893-6080(98)00035-5
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
In this paper, the memory capacity of incompletely connected associative memories is investigated. First, the capacity is derived for memories with fixed parameters. Optimization of the parameters yields a maximum capacity between 0.53 and 0.69 for hetero-association and half of it for autoassociation improving previously reported results. The maximum capacity grows with increasing connectivity of the memory and requires sparse input and output patterns. Further, parameters can be chosen in such a way that the information content per pattern asymptotically approaches 1 with growing size of the memory. (C) 1998 Elsevier Science Ltd. All rights reserved.
引用
收藏
页码:869 / 876
页数:8
相关论文
共 13 条
[1]  
AMARAL DG, 1990, PROG BRAIN RES, V83, P1
[2]   ON SETTING UNIT THRESHOLDS IN AN INCOMPLETELY CONNECTED ASSOCIATIVE NET [J].
BUCKINGHAM, J ;
WILLSHAW, D .
NETWORK-COMPUTATION IN NEURAL SYSTEMS, 1993, 4 (04) :441-459
[3]   INFORMATIONAL CHARACTERISTICS OF NEURAL NETWORKS CAPABLE OF ASSOCIATIVE LEARNING BASED ON HEBBIAN PLASTICITY [J].
FROLOV, AA ;
MURAVEV, IP .
NETWORK-COMPUTATION IN NEURAL SYSTEMS, 1993, 4 (04) :495-536
[4]   Capacity and information efficiency of the associative net [J].
Graham, B ;
Willshaw, D .
NETWORK-COMPUTATION IN NEURAL SYSTEMS, 1997, 8 (01) :35-54
[5]  
HOGAN J, 1995, RANDOM NEURAL NETWOR
[6]  
Kohonen T, 1977, ASS MEMORY
[7]  
LONGUETHIGGINS HC, 1970, REV BIOPHYSIQUE, V3, P223
[9]  
PALM G, 1980, BIOL CYBERN, V36, P19, DOI 10.1007/BF00337019
[10]  
PALM G, 1993, NEURAL ASS MEMORIES