A Robust Morphological Associative Memory Endowed with Dendrites

Morphological neural networks are based on a new paradigm for neural computing. The basic neural computation in a morphological neuron takes the maximum or minimum of the sums of neural values and their corresponding synaptic weights. As a consequence, the properties of morphological neural networks are drastically different than those of traditional neural network models. By making the morphological neuron incorporate dendritic processes, a more realistic model is established. In this paper, we restrict our attention to morphological associative memory endowed with dendrites (MAMED). After a brief review of MAMED and a short discussion about the disadvantages of MAMED in coping with random noises, we present an efficient way of choosing parameter for MAMED taking into account position characteristics of patterns. Our experimental results demonstrate that our way not only makes MAMED be robust in the presence of random noises, but avoids a series of problems brought by choosing arbitrarily