Improved Knowledge Distillation via Teacher Assistant
This addresses a practical limitation in deploying compressed neural networks on edge devices, though it represents an incremental improvement to existing knowledge distillation methods.
The paper tackles the problem of performance degradation in knowledge distillation when there's a large size gap between teacher and student networks, and proposes using intermediate-sized teacher assistant networks to bridge this gap. Experiments on CIFAR-10, CIFAR-100, and ImageNet datasets show the approach effectively improves student network performance.
Despite the fact that deep neural networks are powerful models and achieve appealing results on many tasks, they are too large to be deployed on edge devices like smartphones or embedded sensor nodes. There have been efforts to compress these networks, and a popular method is knowledge distillation, where a large (teacher) pre-trained network is used to train a smaller (student) network. However, in this paper, we show that the student network performance degrades when the gap between student and teacher is large. Given a fixed student network, one cannot employ an arbitrarily large teacher, or in other words, a teacher can effectively transfer its knowledge to students up to a certain size, not smaller. To alleviate this shortcoming, we introduce multi-step knowledge distillation, which employs an intermediate-sized network (teacher assistant) to bridge the gap between the student and the teacher. Moreover, we study the effect of teacher assistant size and extend the framework to multi-step distillation. Theoretical analysis and extensive experiments on CIFAR-10,100 and ImageNet datasets and on CNN and ResNet architectures substantiate the effectiveness of our proposed approach.