An empirical analysis of different sparse penalties for autoencoder in unsupervised feature learning | IEEE Conference Publication | IEEE Xplore