Artificial Neural Networks (ANNs) are broadly used to solve various problems in different. The performance of each ANN highly depends on two factors, i.e. topology and parameters of ANN (typical te weights). There are many differen ANN learning algorithms that change values of parameters leaving, howover, the structure completely intact. The Back - Propagation algorithm is the best -know alogorithm taht belongs to that class. When using such algorithms, the process of searching for a proper ANN topology is the task of ANN designer who arbitrarily chooses ANN structure, activaates learning process, and finally pust ANN to a test.