Optimizing Performance Measures for Feature Selection
Qi Mao, Ivor Wai-Hung Tsang · Proceedings · 2011
Feature selection with specific multivariate performance measures is the key to the success of many applications, such as information retrieval and bioinformatics. The existing feature selection methods are usually designed for classification error. In this paper, we present a unified feature selection framework for general loss functions. In particular, we study the novel feature selection paradigm by optimizing multivariate performance measures. The resultant formulation is a challenging problem for high-dimensional data. Hence, a two-layer cutting plane algorithm is proposed to solve this problem, and the convergence is presented. Extensive experiments on large-scale and high-dimensional real world datasets show that the proposed method outperforms l1-SVM and SVM-RFE when choosing a small subset of features, and achieves significantly improved performances over SVMperfin terms of F1-score.
Cited by 8 works. doi:10.1109/icdm.2011.113
Authors
References
- Gene Selection for Cancer Classification using Support Vector Machines
- The Cutting-Plane Method for Solving Convex Programs
- Training linear SVMs in linear time
- Convex analysis and minimization algorithms
- Large Margin Methods for Structured and Interdependent Output Variables
- Large Scale Multiple Kernel Learning
- A support vector method for multivariate performance measures
- Cutting-plane training of structural SVMs
- A support vector method for optimizing average precision
- Convex Analysis and Nonlinear Optimization