A learning rule for asynchronous perceptrons with feedback in a combinatorial environment