Student Helping Teacher: Teacher Evolution via Self-Knowledge Distillation
This work addresses the challenge of enhancing teacher models for deployment in machine learning, offering a novel approach that is incremental but provides specific performance improvements.
The paper tackles the problem of improving teacher networks in knowledge distillation by introducing a student-helping-teacher framework called TESKD, where the teacher learns from multiple hierarchical students via shared backbones, resulting in ResNet-18 achieving 79.15% accuracy on CIFAR-100 and 71.14% on ImageNet, with gains of 4.74% and 1.43% over baselines.
Knowledge distillation usually transfers the knowledge from a pre-trained cumbersome teacher network to a compact student network, which follows the classical teacher-teaching-student paradigm. Based on this paradigm, previous methods mostly focus on how to efficiently train a better student network for deployment. Different from the existing practices, in this paper, we propose a novel student-helping-teacher formula, Teacher Evolution via Self-Knowledge Distillation (TESKD), where the target teacher (for deployment) is learned with the help of multiple hierarchical students by sharing the structural backbone. The diverse feedback from multiple students allows the teacher to improve itself through the shared feature representations. The effectiveness of our proposed framework is demonstrated by extensive experiments with various network settings on two standard benchmarks including CIFAR-100 and ImageNet. Notably, when trained together with our proposed method, ResNet-18 achieves 79.15% and 71.14% accuracy on CIFAR-100 and ImageNet, outperforming the baseline results by 4.74% and 1.43%, respectively. The code is available at: https://github.com/zhengli427/TESKD.