模型壓縮相關文章
- Learning both Weights and Connections for Efficient Neural Networks (NIPS2015)
- Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding(ICLR2016)
Learning both Weights and Connections for Efficient Neural Networks (NIPS2015)
論文目的:
訓練過程中不僅學習權重參數,也學習網絡連接的重要性,把不重要的刪除掉。
論文內容:
1.使用L2正則化
2.Drop 比率調節
3.參數共適應性,修剪之后重新訓練的時候,參數使用原來的參數效果好。
4.迭代修剪連接,修剪-》訓練-》修剪-》訓練
5.修剪0值神經元
Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding(ICLR2016)
論文內容:
修剪,參數精度變小,哈夫曼編碼 壓縮網絡