Algebraic neural network

A method is described to transform a Boolean function in another in a way to be realised with the McCulloch 's linear neurone. Only few Boolean functions can be realised with McCulloch 's linear neurone. Therefore it will be very useful to know how a Boolean function can be changed in a way to obtain functions that we can realised. The method collect different inputs and gives its the same code (identification map) without enter in contradiction with the Boolean function, i.e., any two inputs with the same code have the same value for the Boolean function. When the Boolean function is linear its weights are the average value of the weights of basic Boolean functions, the threshold is give by the calculated weights. With the parameters of the different identification maps and the linear neurone, the Boolean function is coded. When only parts of the Boolean function are known, we can code the Boolean function and calculate possible values for unknown values of the Boolean function. The learning process by which the parameters are calculated utilise only direct method without any back calculation