A local search method for pattern classification

We present a new method to compute depth-three threshold circuits for pattern classi cation problems. The rst layer of the circuits is calculated from a sample set S of the classi cation problem by a local search strategy that minimises the error on S for each individual gate. The local search is based on simulated annealing with the logarithmic cooling schedule c(k) = = ln (k + 2). The parameter depends on S and the neighbourhood relation is determined by the classical Perceptron algorithm. The approach is applied to the recognition of focal liver tumours.