Application of a fuzzy integral for weak classifiers boosting
A. V. Samorodov · Pattern Recognition and Image Analysis · 2011
A new multiclassifier algorithm, called FuzzyBoost, is proposed. FuzzyBoost provides nonlinear composition model construction and is based on the well-known AdaBoost algorithm, but with additional steps for estimation of fuzzy densities of weak classifiers and calculation of the fuzzy integral instead of the AdaBoost linear aggregation rule at each step of boosting. Experimental studies demonstrated that Fuzzy-Boost has better generalization ability than AdaBoost in the cases of small-size training sets and small-size feature space with correlated features.