Supervised Feature Selection With Orthogonal Regression and Feature Weighting

Effective features can improve the performance of a model and help us understand characteristics underlying structure complex data. Previously proposed feature selection methods usually cannot retain more discriminative information. To address this shortcoming, we propose novel supervised orthogonal least square regression with weighting for selection. The optimization problem objective function be solved by employing generalized power iteration augmented Lagrangian multiplier methods. Experimental results show that method effectively reduce dimensionality obtain better classification than traditional convergence our iterative is also proved. Consequently, effectiveness superiority are verified both theoretically experimentally.