Supervised feature selection using sparse evolutionary training and neuron strength

Feature selection has been used to battle the ever-increasing dimensionality of datasets usedfor machine learning applications. Many feature selection methods, such as the Chi-Square test andLaplacian score, determine feature importance via a hand-crafted metric, often tailored to a specifictype of dataset. This paper proposes a method for deciding feature importance by training a supervisedsparse neural network model usingSparse Evolutionary Trainingand scoring features depending onNeuron Strength. The features are selected in one shot after a network has been trained and, canoutperform Chi-Square test feature selection, performing best in image recognition tasks.