Autoassociative neural memory capacity and dynamics

Storage capacity and retrieval dynamics of autoassociative neural memories (ANMs) are investigated. An associative memory capacity measure based on the memory´s performance versus the number of stored memory vectors (where performance is defined in terms of the memory´s error-correcting ability and its fundamental memories´ attraction volumes) is introduced. The proposed capacity/performance measure has been tested for several ANMs having the same dynamic Hopfield-memory-like architecture, each using a different recording technique. Correlation, generalized inverse (orthogonal), and Ho-Kashyap memory recordings have been investigated. Monte Carlo analysis has been performed on ANMs recorded with randomly generated patterns in order to determine and compare performance characteristics and retrieval dynamics