Artificial neural network for mapping

Summary form only given, as follows. A new learning algorithm is presented for a mapping artificial neural network. The algorithm was discovered during experimentation with backpropagation and counterpropagation networks. The backpropagation network has an excellent way of representing the knowledge, but its learning procedure converges very slowly toward the desired solution. On the other hand, the counterpropagation network´s learning behavior is very predictable and straightforward, but the knowledge acquired is somewhat deficient. The author uses the same knowledge representation as in the counterpropagation network, but the learning algorithm is modified to overcome the described deficiency.<>