立即学习:https://edu.youkuaiyun.com/course/play/28521/394838?utm_source=blogtoedu
v2的改进:
1. Batch normalization:
- no dropout, conv + batch normalization
- regularization for each layer --> easier to 收敛
- now seems 'BN' be the necessary of the network processing
580

被折叠的 条评论
为什么被折叠?



