Extraction of Logical Rules from Neural Networks

A new architecture and method for feature selection and extraction of logical rules from neural networks trained with backpropagation algorithm is presented. The network consists of nodes that discover linguistic features and nodes that discover logical rules. Most weights are constrained to ±1 or zero values. The relevant input features are automatically generated and selected by the network. Rules are generated consecutively, from the most general, covering many training examples, to the most specific, covering exceptions only. Automatic weight pruning ensures that a minimal number of logical rules is found. Results for the Iris classification problem illustrate the efficiency of this method. Keywords Neural networks, MLP, backpropagation, logical rule extraction, feature selection, Iris dataset.

• Type: Journal
• Year: 1998
• Where: NPL
• Authors: Wlodzislaw Duch, Rafal Adamczak, Krzysztof Grabczewski
• Keywords: Logical Rule, Neural Networks, NPL 1998, Relevant Input Features