machine-learning Neural Networks Softmax Function


Softmax regression (or multinomial logistic regression) is a generalization of logistic regression to the case where we want to handle multiple classes. It is particularly useful for neural networks where we want to apply non-binary classification. In this case, simple logistic regression is not sufficient. We'd need a probability distribution across all labels, which is what softmax gives us.

Softmax is computed with the below formula:


___________________________Where does it fit in? _____________________________

Softmax To normalise a vector by applying the softmax function to it with numpy, use:

np.exp(x) / np.sum(np.exp(x))

Where x is the activation from the final layer of the ANN.