Associative memories based on networks of delay differential equations

In this work, a method for storing and retrieving spatio-temporal patterns in large systems of coupled delay differential equations is presented. Spatio-temporal patterns are sets of sequences of binary variables of fixed period that are embedded in the network dynamics as stable limit cycles. As a consequence, an input signal converges to the limit cycle that best represents it. A given set of limit cycles is constructed using a generalization of the correlation learning rule in the definition of the couplings