Summary of Knowledge Distillation Based on Transformed Teacher Matching, by Kaixiang Zheng and En-hui Yang
Knowledge Distillation Based on Transformed Teacher Matching
by Kaixiang Zheng, En-Hui Yang
First submitted to arxiv on: 17 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computer Vision and Pattern Recognition (cs.CV)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper proposes a new technique in knowledge distillation, dubbed transformed teacher matching (TTM), which drops temperature scaling on the student side. This approach adds an inherent Rényi entropy term to the objective function, serving as an extra regularization term. Experimental results show that TTM leads to better generalization and accuracy performance compared to original KD. To further enhance student capabilities, the paper introduces weighted TTM (WTTM), which achieves state-of-the-art accuracy performance. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper develops a new way of sharing knowledge between models called Transformed Teacher Matching. It does this by changing how we use “temperature scaling” in machine learning. This change helps the student model learn better and make fewer mistakes. The researchers tested their idea and found that it works well. They also created an even better version called Weighted TTM, which does even better. |
Keywords
* Artificial intelligence * Generalization * Knowledge distillation * Machine learning * Objective function * Regularization * Student model * Temperature
