Chernoff distance and Relief feature selection

摘要: In classification, a large number of features often make the design classifier difficult and degrades its performance. such situations, feature selection or dimensionality reduction methods play an important role in building classifiers by significantly reducing features. There are many techniques for classification literature. The most popular one is Fisher's linear discriminant analysis (LDA). For two class problems, LDA simply tries to separate means as much possible. multi-class case, does not guarantee capture all relevant information task. To address this problem, problem cast into binary problem. objective becomes find subspace where classes well separated. This formulation only simplifies but also works practice. However, it lacks theoretical justification. We show paper connection between above RELIEF, thereby providing sound basis observed benefits associated with formulation. Experimental results provided that corroborate our analysis.