Leveraging sparse and shared feature activations for disentangled representation learning | Read Paper on Bytez