We have employed information theory techniques to simplify the identification task of ANNs for large and complex databases. These techniques pre-process the spectrum before it is seen by the ANN. One technique that has emerged from this work is called """"""""frequency-information transformation"""""""" (FIT). We have shown that FIT significantly simplifies the classification task for the database of 1H-NMR spectra of xyloglucan oligosaccharides by eliminating the large amount of information contained in 1H-NMR spectra that is irrelevant for classification purposes. However, the technique needs further development in order to be applied to larger databases such as the 1H-NMR spectral database of N-linked oligosaccharides. We have applied for separate NIH funding for this project as an RO1. A manuscript is in preparation describing this work.
Showing the most recent 10 out of 245 publications