An Efficient Feature Selection Algorithm Based on Kernel Function

Abstract

The traditional feature selection algorithms do not consider the nonlinear relationships between features because their objective functions are linear. We propose a feature selection algorithm for capturing nonlinear relationships between features. By introducing a kernel function, the algorithm maps the original data into the high-dimensional kernel space for calculation, then it can consider the relationship among the features. We use L1 norm and L2,1-norm for the regularization factor to double-constrain, not only improves the accuracy of the algorithm but also makes the regularization factor W more sparse. Because two kinds of norms are used simultaneously, they make the objective function difficult to directly derive, this paper designed a superior method to optimize. Experimental results show that the classification performance and stability are stronger than other similar algorithms.