Error tolerant associative memory

Abstract-- This paper surveys two advanced associative memory models[8][5]. The first model was derived from the projection on a closed convex set spanned by patterns. The second model was derived from training weights to improve the error tolerance of the Hopfield network. Both models are designed to resolve the insufficiencies of the Hopfield network. These insufficiencies are loading capacity, limit cycles, and stability with respect to noisy patterns.

• Type: Journal
• Year: 1999
• Where: BC
• Authors: Cheng-Yuan Liou, Shao-Kuo Yuan
• Keywords: Associative Memory Models, BC 1999, Closed Convex, Hopfield Network