Sparse Autoencoder In Deep Learning