Loss

The task for our model is to classify input text sequences as either negative (label 0) or positive (label 1). This is equivalent to binary classification. As with regular binary classification, we use sigmoid ...