Learning from Stochastic Teacher Representations Using Student-Guided Knowledge Distillation
This work addresses efficiency in resource-constrained applications like wearable devices by reducing the need for multiple teacher models in knowledge distillation.
The paper tackles the problem of generating diverse teacher representations for knowledge distillation without training multiple models, by using distillation-time dropout and a student-guided filtering strategy. The proposed method outperforms state-of-the-art approaches on various datasets without increasing model size or computational complexity.
Advances in self-distillation have shown that when knowledge is distilled from a teacher to a student using the same deep learning (DL) architecture, the student performance can surpass the teacher particularly when the network is overparameterized and the teacher is trained with early stopping. Alternatively, ensemble learning also improves performance, although training, storing, and deploying multiple models becomes impractical as the number of models grows. Even distilling an ensemble to a single student model or weight averaging methods first requires training of multiple teacher models and does not fully leverage the inherent stochasticity for generating and distilling diversity in DL models. These constraints are particularly prohibitive in resource-constrained or latency-sensitive applications such as wearable devices. This paper proposes to train only one model and generate multiple diverse teacher representations using distillation-time dropout. However, generating these representations stochastically leads to noisy representations that are misaligned with the learned task. To overcome this problem, a novel stochastic self-distillation (SSD) training strategy is introduced for filtering and weighting teacher representation to distill from task-relevant representations only, using student-guided knowledge distillation (SGKD). The student representation at each distillation step is used as authority to guide the distillation process. Experimental results on real-world affective computing, wearable/biosignal datasets from the UCR Archive, the HAR dataset, and image classification datasets show that the proposed SSD method can outperform state-of-the-art methods without increasing the model size at both training and testing time, and incurs negligible computational complexity compared to state-of-the-art ensemble learning and weight averaging methods.