Feature selection based on linear separability and a CPL criterion function

摘要: Linear separability of data sets is one the basic concepts in theory neural networks and pattern recognition. Data are often linearly separable because their high dimensionality. Such case genomic data, which a small number cases represented space with extremely An evaluation linear two can be combined feature selection carried out through minimisation convex piecewise-linear (CPL) criterion function. The perceptron function belongs to CPL family. basis exchange algorithms allow us find minimal values functions efficiently, even large, multidimensional sets.