Learning Model-Based Sparsity via Projected Gradient Descent | Read Paper on Bytez