Part of Advances in Neural Information Processing Systems 4 (NIPS 1991)
Geoffrey Towell, Jude Shavlik
We propose and empirically evaluate a method for the extraction of expert(cid:173) comprehensible rules from trained neural networks. Our method operates in the context of a three-step process for learning that uses rule-based domain knowledge in combination with neural networks. Empirical tests using real(cid:173) worlds problems from molecular biology show that the rules our method extracts from trained neural networks: closely reproduce the accuracy of the network from which they came, are superior to the rules derived by a learning system that directly refines symbolic rules, and are expert-comprehensible.