Hands-On Machine Learning for Algorithmic Trading

The cross-entropy cost function

The final piece is the cost function to evaluate the neural network output based on the given label. The cost function, J, uses the cross-entropy loss, ξ, which sums the deviations of the predictions for each class, C, from the actual outcome, as follows:

It takes the following form in Python:

def loss(y_hat, y_true):     """Cross-entropy"""     return - (y_true * np.log(y_hat)).sum()