happynear / happynear/caffe-windows
关于全连接层Backward部分中(141~165行代码)weight的求导问题
Open
- Dominant language
- C++
- Stars
- 1.3k
- Forks
- 637
- PR merge metrics
- No merged PRs in 30d
Description
https://github.com/happynear/caffe-windows/blob/2e9ade3075d86321342966ed0aa2961031a06daf/src/caffe/layers/inner_product_layer.cpp#L141-L165
@happynear 你好,既然在Forward中weight进行了归一化:Y=WX/||W||,那么在Backward中weight的导数计算方式怎么跟没有进行归一化(Y=WX)时一样,都是bottom_data*top_diff (for transpose top_diff )或者top_diff *bottom_data(for no transpose top_diff),这样计算是为了工程上处理方便,还是有别的特殊考虑??
Contributor guide
Assessment
This issue has not been assessed yet.