Search for: [Abstract = "Importance of feature selection techniques in multidimensional data analysis is nowadays beyond doubt. It is especially so in such learning tasks which are characterized by a very high dimensionality and a low number of learning examples. An alternative approach to well known and commonly used selection methods \(e.g. backward, forward, stepwise\) is to use the Akaike Information Criterion \(AIC\) for feature selection investigating the whole feature set simultaneously. An experimental approach to feature selection suggested in the paper is based on so\-called AIC Improvement Matrices, which describe the situation in the whole feature set. Besides paying attention to AIC selection algorithms refer also to correlation between features in the data set."]

Number of results: 1