B. Raffin et Mb. Gordon, LEARNING AND GENERALIZATION WITH MINIMERROR, A TEMPERATURE-DEPENDENT LEARNING ALGORITHM, Neural computation, 7(6), 1995, pp. 1206-1224
We study the numerical performances of Minimerror, a recently introduc
ed learning algorithm for the perceptron that has analytically been sh
own to be optimal both on learning linearly and nonlinearly separable
functions. We present its implementation on learning linearly separabl
e boolean functions. Numerical results are in excellent agreement with
the theoretical predictions.