Maybe there is a bug?

[KangGrandesty]

drops_ dose not be reset after Forward and Backward, it still keeps the value of last Forward.
I suggest to reset drops_ at the beginning of Forward.
this is my implementation of FractalNet with global and local drop in caffe:https://github.com/KangGrandesty/fractalnet.
It may be useful for you.

[zhanglonghao1992]

@KangGrandesty Hi,I read your code, its cool but i have one question.
In fractal_join_layer.cpp, I dont understand why "Dtype mult = Dtype(bottom_size)/Dtype(total_undrop_)"
I think Dtype mult = Dtype(1)/Dtype(total_undrop_), that makes the top is element wise mean of the bottom

[KangGrandesty]

@zhanglonghao1992 I just set it like the dropout layer. if half of neurons are dropout randomly, the rest will be scale to sure the output has a closer expected value with non-dropout. Then, I change it to element-wise mean and modify the code.