Knowledge distillation in wide neural networks: risk bound, data efficiency and imperfect teacher

Ji, Guangda and Zhu, Zhanxing (2020) Knowledge distillation in wide neural networks: risk bound, data efficiency and imperfect teacher. Larochelle, H., Ranzato, M., Hadsell, R., Balcan, M.F. and Lin, H. (eds.) In Advances in Neural Information Processing Systems 33. Neural Information Processing Systems Foundation. 11 pp .

Abstract

Knowledge distillation is a strategy of training a student network with guide of the soft output from a teacher network. It has been a successful method of model compression and knowledge transfer. However, currently knowledge distillation lacks a convincing theoretical understanding. On the other hand, recent finding on neural tangent kernel enables us to approximate a wide neural network with a linear model of the network’s random features. In this paper, we theoretically analyze the knowledge distillation of a wide neural network. First we provide a transfer risk bound for the linearized model of the network. Then we propose a metric of the task’s training difficulty, called data inefficiency. Based on this metric, we show that for a perfect teacher, a high ratio of teacher’s soft labels can be beneficial. Finally, for the case of imperfect teacher, we find that hard labels can correct teacher’s wrong prediction, which explains the practice of mixing hard and soft labels.

This record has no associated files available for download.

Contributors

Download statistics