Adaptive learning rule for binary couplings networks

This paper presents a new adaptive iterative learning rule for binary couplings networks. Unlike previous approaches, the algorithm adapts to pattern correlations during learning and succeeds to store highly correlated patterns. Also, by supplying a set of default stabilities to the learning rule, the recall properties of the network can be adjusted for each pattern. Simulations results of pattern recall in a simple recursive network demonstrate the storage and associative memory properties of the trained network and show the advantage over older learning rules. Note that the adaption step of the learning rule can also be applied to other learning algorithms. Applications to multi-layer networks and hardware implementation are discussed.