Binary Kernel in Morphological Associative Memory

Abstract

This work presents a binary implementation of the internal representa- tion kernel in morphological associative memory (MAM) [IEEE International Con- ference on System, Man, and Cybermetics 4: 3570–3575, 1997]. This kernel was proposed by Ritter et al. [IEEE International Conference on System, Man, and Cy- bermetics 4: 3570–3575, 1997] and further solved by Hattori et al. [Proceedings of the 9th International Conference on Neural Information Processing 2:1058–1063, 2002]. Ritter used a trial and error method to search for a suitable kernel Z for MAM and attempted to work out a theoretical way to estimate the kernel. Hattori proposed a fast solution for finding this kernel. This work shows that in binary cases, when solving the kernel, only those bits satisfying certain specific condi- tions need to be checked. This can speed the checking procedure in large scale data processing.