In the code, loaded samples seem only to substract mean but not to be normalized, which leads to very large loss. Why not normalize the samples?
In the code, loaded samples seem only to substract mean but not to be normalized, which leads to very large loss. Why not normalize the samples?