A New Method of Morphological Associative Memories

摘要

The morphological associative memories (MAM) have many attractive advantages. However, they can not give a guarantee that morphological he-tero-associative memories are perfect, even if input patterns are perfect. In addition, the problem with the associative memory matrixes W_(XY) and M_(XY) is that W_(XY) is incapable of handling dilative noise while M_(XY) is incapable of effectively handling erosive noise. In this paper, the new methods of MAM, ~+W_(XY) and ~+M_(XY) are proposed. The certain qualifications that make ~+W_(XY) and ~+M_(XY) be perfect memories are analyzed and proved. As far as the hetero-associative memories are concerned, although ~+W_(XY) and ~+M_(XY) are not perfect, they are complements to original W_(XY) and M_(XY)- ~+W_(XY) is capable of handling dilative noise while ~+M_(XY) is capable of effectively handling erosive noise. Therefore they can be put together with original W_(XY) and M_(XY) to learn from others' strong points to offset ones' own weakness and to make the effect of hetero-associative memories and pattern recognition better. The calculation results demonstrate that both ~+W_(XY) and ~+M_(XY) are effectual.