Excuse me, I just don't understand why the derivative function of SoftMax is np.ones_like(x).... $$ $S(x_{i}) = \frac{e^{x_{i}}}{\sum{x_{i}}} $ $$