Binary cross-entropy loss
The binary cross-entropy loss is given by:
for training examples (indexed by ) where is the class label (0 or 1) and is the prediction for that example (i.e. the predicted probability that it is a positive example). Thus is the probability that it is a negative example.