Coarse pruning of convolutional neural networks with random masks
這是ICLR 2017的一篇文章，文章認爲既然我們無法直觀得去衡量weights，layers，kernels的重要性，那我們就用random的方式。對於N個要剪枝的對象，我們可以有2^N種組合。在給定裁剪率 α 的情況下，就有 α*2^N種組合。從這些組合中挑選在驗證集上精度最大的作爲局部最優剪枝。文章建議的N=50，α=40%。 單純從效果上來講，還不錯： 討論 但是，讀者從自己做裁剪工作的角
相關文章
- 1. Convolutional Neural Networks---Foundations of Convolutional Neural Networks
- 2. Bag of Tricks for Image Classification with Convolutional Neural Networks
- 3. Bag of Tricks for Convolutional Neural Networks
- 4. Pruning Convolutional Neural Networks For Resource Efficient Inference
- 5. Pruning convolutional neural networks for resource efficent inference
- 6. Designing Energy-Efficient Convolutional Neural Networks using Energy-Aware Pruning
- 7. IMAGE DENOISING WITH GRAPH-CONVOLUTIONAL NEURAL NETWORKS
- 8. Convolutional Neural Networks: Application
- 9. ImageNet Classification with Deep Convolutional Neural Networks
- 10. Convolutional Neural Networks with Alternately Updated Clique
- 更多相關文章...
- • XSLT
- • XSLT
- • 爲了進字節跳動，我精選了29道Java經典算法題，帶詳細講解
- • RxJava操作符（一）Creating Observables
相關標籤/搜索
每日一句
-
每一个你不满意的现在，都有一个你没有努力的曾经。
歡迎關注本站公眾號,獲取更多信息
相關文章
- 1. Convolutional Neural Networks---Foundations of Convolutional Neural Networks
- 2. Bag of Tricks for Image Classification with Convolutional Neural Networks
- 3. Bag of Tricks for Convolutional Neural Networks
- 4. Pruning Convolutional Neural Networks For Resource Efficient Inference
- 5. Pruning convolutional neural networks for resource efficent inference
- 6. Designing Energy-Efficient Convolutional Neural Networks using Energy-Aware Pruning
- 7. IMAGE DENOISING WITH GRAPH-CONVOLUTIONAL NEURAL NETWORKS
- 8. Convolutional Neural Networks: Application
- 9. ImageNet Classification with Deep Convolutional Neural Networks
- 10. Convolutional Neural Networks with Alternately Updated Clique