Supporting compressed-sparse activations and weights on SIMD-like accelerator for sparse convolutional neural networks | IEEE Conference Publication | IEEE Xplore