INVESTIGATION OF INFRARED SPECTRA-STRUCTURE CORRELATION USING KOHONENAND COUNTERPROPAGATION NEURAL-NETWORK

Two different artificial neural networks (ANNs) for infrared spectra a nalysis are presented: the self-organizing Kohonen ANN for mapping of the infrared spectra into a 2-D plane and the counterpropagation ANN f or determination of the structural features of organic compounds based on their infrared spectra. The preliminary learning in the Kohonen AN N with all spectra from the collection yields the information of possi ble grouping. The preliminary grouping has been used for the separatio n of spectra into the training and into the test set containing 755 an d 2529 ''spectrum-structure'' pairs, respectively. The counterpropagat ion ANN trained on the ''spectrum-structure'' pairs from the training set has the ability to predict, with an average prediction ability of 0.77 and an average reliability of 0.82, structural fragments of an un known compound from its infrared spectrum. Additionally, the counterpr opagation ANN offers the possibility to simulate the infrared spectra from the structure representation.