A0580
Title: Regularized classification with its application to biomedical spectroscopic data
Authors: Ying Zhu - National Institute of Education, Nanyang Technological University (Singapore) [presenting]
Augustine Tuck Lee Tan - Nanyang Technological University (Singapore)
Wai Kwong Cheang - Nanyang Technological University (Singapore)
Abstract: High-dimensional spectroscopic data consist of many overlapping absorption bands sensitive to the physical and chemical states of compounds. Often, only a small subset of spectral features is found essential. Direct implementation of linear discriminant methods on high-dimensional spectroscopic data provides poor classification results due to singularity problem and highly correlated spectral features. A regularized classification model incorporating complex spectral correlation structure enabled an automatic selection of a small number of informative spectral absorption bands for the purpose of classification and interpretation. This model has been applied on biomedical spectroscopic data. The well-performed selection of informative spectral features leads to substantial reduction in model complexity, improvement of classification accuracy, and is particularly helpful for providing us insights in the interpretation of the complex spectroscopic data regarding its active ingredients.
