How to Train the Teacher Model for Effective Knowledge Distillation
This work addresses a specific bottleneck in knowledge distillation for machine learning practitioners, offering an incremental improvement over existing methods.
The paper tackles the problem of improving knowledge distillation by training the teacher model to better estimate the true Bayes conditional probability density, showing that using MSE loss instead of cross-entropy loss boosts student accuracy by up to 2.6% in experiments.
Recently, it was shown that the role of the teacher in knowledge distillation (KD) is to provide the student with an estimate of the true Bayes conditional probability density (BCPD). Notably, the new findings propose that the student's error rate can be upper-bounded by the mean squared error (MSE) between the teacher's output and BCPD. Consequently, to enhance KD efficacy, the teacher should be trained such that its output is close to BCPD in MSE sense. This paper elucidates that training the teacher model with MSE loss equates to minimizing the MSE between its output and BCPD, aligning with its core responsibility of providing the student with a BCPD estimate closely resembling it in MSE terms. In this respect, through a comprehensive set of experiments, we demonstrate that substituting the conventional teacher trained with cross-entropy loss with one trained using MSE loss in state-of-the-art KD methods consistently boosts the student's accuracy, resulting in improvements of up to 2.6\%.