Conceptual

Cross-Entropy Loss

Measuring the gap between a predicted probability distribution and the true outcome; negative log-likelihood as the training loss for classification and language modeling.