2001 Midwest Symposium on Circuits and Systems (MWSCAS 2001), Dayton, OH, United States Of America, 14 - 17 August 2001, vol.2, pp.713-716
We propose a binary associative memory design method to be applied to a class of dynamical neural networks. The method is based on introducing the memory vectors as maximal independent sets to an undirected graph and on designing a dynamical network in order to find a maximal independent set whose characteristic vector is close to the given distorted vector. We show that our method provides the attractiveness for each memory vector and avoids the occurance of spurious states whenever the set of given memory vectors satisfies certain compatibility conditions. We also analyze the application of this design method to the discrete Hopfield network.