Survey and critique of techniques for extracting rules from trained artificial neural networks

摘要: It is becoming increasingly apparent that, without some form of explanation capability, the full potential trained artificial neural networks (ANNs) may not be realised. This survey gives an overview techniques developed to redress this situation. Specifically, focuses on mechanisms, procedures, and algorithms designed insert knowledge into ANNs (knowledge initialisation), extract rules from (rule extraction), utilise refine existing rule bases refinement). The also introduces a new taxonomy for classifying various techniques, discusses their modus operandi, delineates criteria evaluating efficacy.