Neural methods of knowledge extraction

Contrary to the common opinion, neural networks may be used for knowledge e xtraction. Recently, a new methodology of logical rule extraction, optimiza tion and application of rule-based systems has been described, C-MLP2LN alg orithm, based on constrained multilayer perceptron network, is described he re in details and the dynamics of a transition from neural to logical syste m illustrated. The algorithm handles real-valued features, determining appr opriate linguistic variables or membership functions as a part of the rule extraction process. Initial rules are optimized by exploring the accuracy/s implicity tradeoff at the rule extraction stage and the one between reliabi lity of rules and rejection rate at the optimization stage. Gaussian uncert ainties of measurements are assumed during application of crisp logical rul es, leading to "soft trapezoidal" membership functions and allowing to opti mize the linguistic variables using gradient procedures. Comments are made on application of neural networks to knowledge discovery in the benchmark a nd real life problems.