Enhanced Fuzzy Autoassociative Morphological Memory for Binary Pattern Recall

G. Urcid, J.A. Nieves-Vázquez, and C.A. Reyes-García (Mexico)


Binary pattern recognition, fuzzy neural networks, Hamming network, minimax algebra, morphological associative memories.


Autoassociative morphological memories (AMMs) are a class of artificial feedforward neural networks whose computation at each neurode is based on lattice algebra. In a similar way as the classic correlation encoding used for binary patterns in linear associative memories or recurrent content addressable memories such as the Hopfield network, storage and recall in AMMs is also realized using matrix transforms which in the present case correspond to minimax matrix operations. This paper describes an enhanced fuzzy autoassociative morphological memory that couples a fuzzy AMM to a Hamming network that increases the capability of perfect recalls from noisy binary inputs.

Important Links:

Go Back