Abstract: |
Least-SquaresXCSF is an extension of XCS in which classifier prediction iscomputed as a linear combination of classifier inputs and aweight vector associated to each classifier. XCSF can adjustthe weight vector of classifiers to evolve accurate piecewiselinear approximations of functions. The Widrow-Hoff rule,used to update the weight vectors, prevents (when someconditions hold) XCSF from exploiting the expected piecewiselinear approximation. In this paper we replace theWidrow-Hoff rule with linear least-squares and we show thatwith this improvement XCSF can fully exploit its generalizationcapabilities. |
