Application of Fuzzy Neural Network to Spectrum Identification

Abstract

The main concept of using neural network models for pattern association and pattern classification is outlined. Applications for handling spectroscopic patterns are demonstrated with storing UV-spectra recorded under different experimental conditions in a neural network and identifying unknown spectra by presenting them to the network. Essentially Kosko’s Adaptive Bidirectional Associative Memory [2] is explored that enables to learn similar spectra of one chemical compound as well as spectra of different chemical identity. The final comparison of sample and reference spectra is carried out by fuzzy set operations.