The problem of binary channel equalization can be regarded as the problem of dividing a (hyper)plane in two decision regions. In general, the optimum boundary between the decision regions is nonlinear, even if the transmission channel is linear. In this paper we determine a suboptimal piecewise linear decision region boundary from a set of known examples through the application of genetic algorithms. The paper focuses in particular on the adaptation of the basic genetic algorithm machinery to the specific problem at hand, and on the translation of the usual concepts of cross-over and mutation in the geometric domain of application