Extraction of Logical Rules from Neural Networks

Authors:
Włodzisław Duch;Rafał Adamczak;Krzysztof Grąbczewski
Affiliations:
Department of Computer Methods, Nicholas Copernicus University, Grudziądzka 5, 87–100 Toruń, Poland;Department of Computer Methods, Nicholas Copernicus University, Grudziądzka 5, 87–100 Toruń, Poland;Department of Computer Methods, Nicholas Copernicus University, Grudziądzka 5, 87–100 Toruń, Poland
Venue:
Neural Processing Letters
Year:
1998

Citing 1
Cited 5

A practical Bayesian framework for backpropagation networks

Neural Computation

A methodology to explain neural network classification

Neural Networks
Observational Logic Integrates Data Mining Based on Statistics and Neural Networks

PKDD '00 Proceedings of the 4th European Conference on Principles of Data Mining and Knowledge Discovery
Extraction of Logical Rules from Data by Means of Piecewise-Linear Neural Networks

DS '02 Proceedings of the 5th International Conference on Discovery Science
Piecewise-Linear Neural Networks and Their Relationship to Rule Extraction from Data

Neural Computation
Evolutionary multi-objective optimization for simultaneous generation of signal-type and symbol-type representations

EMO'05 Proceedings of the Third international conference on Evolutionary Multi-Criterion Optimization

Quantified Score

Hi-index	0.00

Visualization

Abstract

Three neural-based methods for extraction of logical rules from data are presented. These methods facilitate conversion of graded response neuralnetworks into networks performing logical functions. MLP2LN method tries toconvert a standard MLP into a network performing logical operations (LN).C-MLP2LN is a constructive algorithm creating such MLP networks. Logicalinterpretation is assured by adding constraints to the cost function,forcing the weights to ±1 or 0. Skeletal networks emergeensuring that a minimal number of logical rules are found. In both methodsrules covering many training examples are generated before more specificrules covering exceptions. The third method, FSM2LN, is based on theprobability density estimation. Several examples of performance of thesemethods are presented.