We describe and present results of a high-order double layer associative memory (HODLAM). The memory can utilise up to neat-4, 100% of the network storage capacity, has high tolerance to patterns with high cross-correlations and has high potential for translation-invariant pattern recognition. We demonstrate the capability of the memory by using it to recognise a set of 64 characters, each being drawn on an 8X8 pirelated array. We also propose a possible optical implementation scheme for the memory and examine modifications that lead to translation invariance.
David T.Y. Lam,
John E. Carroll,
"Pattern Recognition Using A Double Layer Associative Memory", Proc. SPIE 1154, Real-Time Signal Processing XII, (6 December 1989); doi: 10.1117/12.962383; https://doi.org/10.1117/12.962383