Title :
Effect of pruning small weights in correlation associative memories
Author :
Kothari, Ravi ; Lotlikar, Rohit
Author_Institution :
Dept. of Electr. & Comput. Eng. & Comput. Sci., Cincinnati Univ., OH, USA
Abstract :
In this paper we examine the effect of pruning small weights in correlation associative memories. The effect is investigated by considering a fully interconnected matrix and removing weights satisfying |wij|⩽ε. We show analytically that under suitable constraints on ε, the capacity of a sparsely connected associative memory is comparable to that of a fully connected one, without compromising the basins of attraction around the prototype fixed points. Simulation results supporting the analysis are also presented
Keywords :
Hebbian learning; associative processing; content-addressable storage; matrix algebra; neural nets; optimisation; probability; Hebbian learning; attraction basin; attractor neural networks; correlation associative memory; interconnected matrix; optimisation; probability; small weight pruning; Analytical models; Associative memory; Computer science; Image restoration; Image retrieval; Image storage; Laboratories; Neural networks; Neurons; Prototypes;
Conference_Titel :
Neural Networks,1997., International Conference on
Conference_Location :
Houston, TX
Print_ISBN :
0-7803-4122-8
DOI :
10.1109/ICNN.1997.616200
