AUTOMATED FEATURE-SELECTION WITH A DISTINCTION SENSITIVE LEARNING VECTOR QUANTIZER

An extended version of Kohonen's Learning Vector Quantization (LVQ) al gorithm, called Distinction Sensitive Learning Vector Quantization (DS LVQ), is introduced which overcomes a major problem of LVQ, the depend ency on proper pre-processing methods for scaling and feature selectio n. The algorithm employs a weighted distance function and adapts the m etric with learning. Highest weights are assigned to components in the input vectors which are most informative for classification; non-info rmative components are discarded. The algorithm is applied to the anal yses of multi-channel EEG data and compared with experienced methods.