| Abstract
| - Combining gas phase infrared (IR) spectra with mass spectral (MS)data, a neural network has been developedto predict 26 different molecular substructures from multispectralinformation. The back-propagationprocedure has been used for training, including its previouslypublished modification, the flashcard algorithm.Present functional groups have been detected correctly in 86.4%of all cases, compared with 88.4% usingonly IR and 78.2% using only MS data for training and prediction.For only 8 out of the 26 functionalitiesdoes the joint utilization of infrared and mass spectra yield betterprediction results, with the greatestimprovement being for halogen bond predictions. The prediction offunctional group absence results inaccuracy of about 95.5% for both IR and IR/MS networks but only 87.1%for a stand alone MS network.Insights have been gained into the suitability of both data setsfor neural network training by presenting justIR or MS data to a jointly trained neural network, revealing the amountof information the network utilizesfrom either spectroscopic technique. In addition, an algorithmwhich produces balanced training and testsets for multi-output neural networks has been devised.
|
