Unsupervised Feature Selection Using Rbf Autoencoder

In this paper, a novel learning approach to solve unsupervised feature selection in high-dimensional data is proposed, namely Radial Basis Function Autoencoder feature selection (RAFS). This method based on autoencoder uses the radial basis function to achieve mapping instead of the weight. We also consider penalty to give a powerful constraint on redundant features. In extensive experiments, our method shows its outperformance in fair comparison with several other methods.