PERFORMANCE ANALYSIS OF A CONVERGED SINGLE-LAYER PERCEPTRON FOR NONSEPARABLE DATA MODELS WITH BIAS TERMS

Rosenblatt's algorithm is a recursive method used to adjust the weight s of a single-layer perceptron. It is capable of partitioning the inpu t. signal space into two regions that are separated by a hyperplane bo undary. Thus, when the values of the input signal are linearly separab le, the algorithm will converge to a stable stationary point that yiel ds zero mean-square error. In this paper, we examine the stationary po ints of Rosenblatt's algorithm when the data is not linearly separable . A system identification model is used to generate the data. The mode l incorporates the effects of bias terms so that the hyperplane bounda ries do not necessarily pass through the origin of the signal space. A n expression, is also derived for the probability of an incorrect clas sification of the output signal when the weights are converged at a st ationary point.