Zebra: Memory Bandwidth Reduction for CNN Accelerators with Zero Block Regularization of Activation Maps

The large amount of memory bandwidth between local buffer and external DRAM has become the speedup bottleneck of CNN hardware accelerators, especially for activation maps. To reduce memory bandwidth, we propose to learn pruning unimportant blocks dynamically with zero block regularization of activation maps (Zebra). This strategy has low computational overhead and could easily integrate with other pruning methods for better performance. The results show that the proposed method for Resnet-20 on Tiny-Imagenet can reduce 70% of memory bandwidth and further improve to 76% with the combination of Network Slimming, all within 1% accuracy drops.