模型压缩相关文章
- 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)
论文目的:
训练过程中不仅学习权重参数,也学习网络连接的重要性,把不重要的删除掉。
论文内容:
1.使用L2正则化
2.Drop 比率调节
3.参数共适应性,修剪之后重新训练的时候,参数使用原来的参数效果好。
4.迭代修剪连接,修剪-》训练-》修剪-》训练
5.修剪0值神经元
Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding(ICLR2016)
论文内容:
修剪,参数精度变小,哈夫曼编码 压缩网络