Maximized mutual information using macrocanonical probability distributions

A maximum entropy formulation leads to a neural network which is factorable in both form and function into individual neurons corresponding to the Hopfield neural model. A maximized mutual information criterion dictates the optimal learning methodology using locally available information.