Rethinking Knowledge Distillation: A Data Dependent Regulariser With a Negative Asymmetric Payoff

Open in new window