Deep Neural Network Compression with Emphasis on Sparsification Techniques