Design and analysis of hardware friendly pruning algorithms to accelerate deep neural networks at the edge