A comparison of criterion functions for a neural network applied to binary detection

In this paper, the performance of a neural binary detector for five different criterion functions is analyzed, showing that a typical backpropagation algorithm that uses least-mean-squares (LMS) criterion function, despite of its widely use, is far from achieving the best solution for this problem. By evaluating the detection performance of each network, for the same structure, it is shown how the change of the criterion function improves significantly the solution achieved by the typical LMS error criterion