A Boolean neural network controlling task sequences in a noisy environment

The classical exclusive-or problem and many others like it cannot be performed with networks without hidden units, with which they create their own internal representation of the input patterns. The idea to use multi-layered network for solving that kind of problems has been successful, but it requires to find powerful learning rules for networks with hidden units, that are also very simple guaranteed learning rules. We propose a quite general solution for implementing a Hebbian rule into non-layered Boolean neural network, in order to solve that kind of problems. We also present some experimental results.