Domanda di colloquio di Sony

Self-supervised learning, l1-l2 regularization and their differences, model compression techniques.