NONMONOTONIC BEHAVIOR OF THE CAPACITY IN PHASOR NEURAL NETWORKS

The stochastic dynamics of Q-phasor neural networks is discussed using a probabilistic approach. For layered feedforward architectures and Hebbian learning, exact evolution equations are given for arbitrary Q at both zero and finite temperatures. The capacity-temperature diagram is presented. At zero temperature a nonmonotonic behavior of the capacity is found as a function of the number of phases Q, contrary to other multistate neural network models. © 1994 The American Physical Society.