Feature-Level Ensemble Knowledge Distillation for Aggregating Knowledge from Multiple Networks

Abstract

Knowledge Distillation (KD) aims to transfer knowledge in a teacher-student framework, by providing the predictions of the teacher network to the student network in the training stage to help the student network generalize better. It can use either a teacher with high capacity or an ensemble of multiple teachers. However, the latter is not convenient when one wants to use feature-map-based distillation methods. In this paper, we empirically show that using several non-linear transformation layer cope well with multiple-teacher setting compared to other kinds of feature-map-level distillation methods. Comprehensively, this paper proposes a versatile and powerful training algorithm named FEature-level Ensemble knowledge Distillation (FEED), which aims to transfer the ensemble knowledge using multiple teacher networks. In this study, we introduce a couple of training algorithms that transfer ensemble knowledge to the student at the feature-map-level. Among the feature-map-level distillation methods, using several non-linear transformations in parallel for transferring the knowledge of the multiple teachers helps the student find more generalized solutions. We name this method as parallel FEED, and experimental results on CIFAR-100 and ImageNet show that our method has clear performance enhancements, without introducing any additional parameters or computations at test time. We also show the experimental results of sequentially feeding teacher's information to the student, hence the name sequential FEED, and discuss the lessons obtained. Additionally, the empirical results on measuring the reconstruction errors at the feature map give hints for the enhancements.