Rules Extraction from Neural Network Based on Immune Programming Algorithm

From the functional point of view, a method that employed an Immune Programming algorithm to extract rules from Neural Network (NN) was presented. According to quantifying color feature of landscape images had been marked with emotion values in China Affective Picture System (CAPS), the method clustered outputs of hidden neurons of NN, reduced the searching scale, and the mapping between color and subjective impressions was established. The result of experiment shows it can extract reasonably succinct sign rules that have high predictive accuracy rates.