模型壓縮備用

categorized into four schemes: parameter pruning and sharing low-rank factorization transferred/compact convolutional filters knowledge distillation Achieving compacting and accelerating CNNs model ca

>>阅读原文<<