 |
Bilevel optimization for feature selection in binary classification with support vector machines
1 : SKEMA Business School
Skema Business School
We propose a bilevel optimization formulation for feature selection in binary classification tasks, where the classifier is an support vector machines (SVM) with a linear kernel. Our formulation allows us to integrate both the training and testing phases of classification. In our bilevel model, the leader decides the features to be selected in order to maximize the classification accuracy, and the follower (i.e. the SVM) determines the optimal parameters of the separating hyperplane given the features selected by the leader.
