“…As a consequence, careful examination of the hit list produced by the library search algorithm is absolutely essential, and in many cases it remains necessary for the analyst to interpret the mass spectra in order to elucidate an unknown structure. In these circumstances the use of ANN, combined with other taxonomic methods, can provide valuable assistance in the structure determination. − Although the application of ANN to identify specific SFs, fragments, or substructures from spectroscopic information can be routinely carried out on a microcomputer, subject to the availability of specific software, , the prior construction, training, and testing of ANN requires more powerful hardware and software resources. Furthermore presently there is no well established methodology for ANN development in real practical problems, consequently one has to adopt an empirical approach to choose the architecture, learning method, and input and output vectors which best fit the problem at hand.…”