STATICS AND DYNAMICS OF AN ASHKIN-TELLER NEURAL-NETWORK WITH LOW LOADING

An Ashkin-Teller neural network, allowing for two types of neurons is considered in the case of low loading as a function of the strength of the respective couplings between these neurons. The storage and retri eval of embedded patterns built from the two types of neurons, with di fferent degrees of (in)dependence is studied. In particular, thermodyn amic properties including the existence and stability of Mattis states are discussed. Furthermore, the dynamic behaviour is examined by deri ving flow equations for the macroscopic overlap. It is found that for linked patterns the model shows better retrieval properties than a cor responding Hopfield model.