ON THE CONVERGENCE OF A GROWING TOPOLOGY NEURAL ALGORITHM

A convergence theorem for the Cascade Correlation is described and bas ed on this, results on convergence speed of the algorithm are obtained . It is possible to forecast the cost when adding hidden units: this a llows us to understand if the Cascade Correlation works properly. A cr itical test is presented where the growing process must be modified in order to avoid unsatisfactory outcome; a parameter for detecting such occurrences is proposed and therapies are suggested. The results obta ined in this paper may be used as a kernel for building new types of g rowing networks, whose topology may be better matched with the trainin g data.