softmax_layer

$\frac{\partial L}{\partial x_i}=r[1-\frac{\sum{\hat{p_k}I(l_k=1)}}{\sum{I(l_k=1)}}]^{r-1} \frac{I(l_i=1)}{\sum{I(l_k=1)}}[\sum_k{p_klog(\hat{p_k})+(1-p_k)log(1-\hat{p_k})}]\hat{p_i}(1-\hat{p_i})-[1-\frac{\sum{\hat{p_k}I(l_k=1)}}{\sum{I(l_k=1)}}]^{r}(p_i-\hat{p_i})$

\frac{\partial L}{\partial x_i}=r[1-\frac{\sum{\hat{p_k}I(l_k=1)}}{\sum{I(l_k=1)}}]^{r-1} \frac{I(l_i=1)}{\sum{I(l_k=1)}}[\sum_k{p_klog(\hat{p_k})+(1-p_k)log(1-\hat{p_k})}]\hat{p_i}(1-\hat{p_i})-[1-\frac{\sum{\hat{p_k}I(l_k=1)}}{\sum{I(l_k=1)}}]^{r}(p_i-\hat{p_i})

softmax_loss_layer

caffe cpu 模式
http://blog.csdn.net/zyb19931130/article/details/53842791
gcc要求4.9
不能有中文路径

2015 CVPR

http://blog.csdn.net/ycheng_sjtu/article/details/38535045

图像语义理解：
http://personal.psu.edu/jol2/pub/alipr06.pdf

https://lvyou.baidu.com/notes/f840db1fabb7e03a496792ff?sid=c2fa4653ba3ece0ae6bc95f3&from=search&request_id=2226731513&idx=0

解忧杂货店铺读后感

解忧杂货店铺读后感