Function cross_entropy

11_softmax_and_crossentropy.py:33–37

(actual, predicted)

# whose output is a probability value between 0 and 1. 
# -> loss increases as the predicted probability diverges from the actual label
def cross_entropy(actual, predicted):
    EPS = 1e-15
    predicted = np.clip(predicted, EPS, 1 - EPS)
    loss = -np.sum(actual * np.log(predicted))
    return loss # / float(predicted.shape[0])

# y must be one hot encoded
# if class 0: [1 0 0]