Adaptation of Boolean networks using back-error propagation

摘要

Summary form only given, as follows. The back-error-propagationnmethod has been adapted to give a training procedure for networks ofnBoolean units. The method has been applied to pyramidal networks ofnunits, and simulation studies of the method have been undertaken.nSolutions have been found for hard problems, including the 8-bit paritynproblem. The results compare favorably with those obtained for networksnof semilinear units, but also show the difficulties of applyingnoptimization methods where large plateaus are present in the errornfunction. The method is compared with adaptive procedures for trainingnnetworks of probabilistic logic neurons