Asymmetries in synaptic connections and the nonlinear fokker-planck formalism

Abstract

In previous work we have developed illustrative, neurocomputational models to describe mechanisms associated with mental processes. In these efforts, we have considered mental processes in phenomena such as neurosis, creativity, consciousness/unconsciousness, and some characteristics of the psychoses. Memory associativity is a key feature in the theoretical description of these phenomena, and much of our work has focused on modeling this mechanism. In traditional neural network models of memory, the symmetry of synaptic connections is a necessary condition for reaching stationary states. The assumption of symmetric weights seems however to be biologically unrealistic. Efforts to model stationary network states with asymmetric weights are mathematically complex and are usually applied to restricted situations. This has motivated us to explore the possibility of a new approach to the synaptic symmetry problem, based on its analogies with some features of the nonlinear Fokker-Planck formalism.