Preparing Lessons: Improve Knowledge Distillation with Better Supervision

Abstract

Knowledge distillation (KD) is widely used for training a compact model with the supervision of another large model, which could effectively improve the performance. Previous methods mainly focus on two aspects: 1) training the student to mimic representation space of the teacher; 2) training the model progressively or adding extra module like discriminator. Knowledge from teacher is useful, but it is still not exactly right compared with ground truth. Besides, overly uncertain supervision also influences the result. We introduce two novel approaches, Knowledge Adjustment (KA) and Dynamic Temperature Distillation (DTD), to penalize bad supervision and improve student model. Experiments on CIFAR-100, CINIC-10 and Tiny ImageNet show that our methods get encouraging performance compared with state-of-the-art methods. When combined with other KD-based methods, the performance will be further improved.

References in corpus (12)

• Distilling the Knowledge in a Neural Network
• Paying More Attention to Attention: Improving the Performance of Convolutional Neural Networks via Attention Transfer
• Do Deep Nets Really Need to be Deep?
• Like What You Like: Knowledge Distill via Neuron Selectivity Transfer
• Label Refinery: Improving ImageNet Classification through Label Progression
• Deep Model Compression: Distilling Knowledge from Noisy Teachers
• Knowledge Distillation in Generations: More Tolerant Teachers Educate Better Students
• Training Shallow and Thin Networks for Acceleration via Knowledge Distillation with Conditional Adversarial Networks
• Adaptive Regularization of Labels
• Revisiting Knowledge Distillation via Label Smoothing Regularization
• An Embarrassingly Simple Approach for Knowledge Distillation
• Learning Student Networks via Feature Embedding

Cited by in corpus (6)

• Rethinking Soft Labels for Knowledge Distillation: A Bias-Variance Tradeoff Perspective
• Distilling Knowledge by Mimicking Features
• TextBrewer: An Open-Source Knowledge Distillation Toolkit for Natural Language Processing
• Robustness and Diversity Seeking Data-Free Knowledge Distillation
• KD-Lib: A PyTorch library for Knowledge Distillation, Pruning and Quantization
• Isotonic Data Augmentation for Knowledge Distillation