AN ASSOCIATIVE MEMORY THAT CAN FORM HYPOTHESES - A PHASE-CODED NEURAL-NETWORK

Nonlinear associative memories as realized, e.g., by Hopfield nets are characterized by attractor-type dynamics. When fed with a starting pa ttern, they converge to exactly one of the stored patterns which is su pposed to be most similar. These systems cannot render hypotheses of c lassification, i.e., render several possible answers to a given classi fication problem. Inspired by von der Malsburg's correlation theory of brain function, we extend conventional neural network architectures b y introducing additional dynamical variables. Assuming an oscillatory time structure of neural firing, i.e., the existence of neural clocks, we assign a so-called phase to each formal neuron. The phases explici tly describe detailed correlations of neural activities neglected in c onventional neural network architectures. Implementing this extension into a simple self-organizing network based on a feature map, we prese nt an associative memory that actually is capable of forming hypothese s of classification.