Summary of Epsd: Early Pruning with Self-distillation For Efficient Model Compression, by Dong Chen et al.
EPSD: Early Pruning with Self-Distillation for Efficient Model Compression
by Dong Chen, Ning Liu, Yichen Zhu, Zhengping Che, Rui Ma, Fachao Zhang, Xiaofeng Mou, Yi Chang, Jian Tang
First submitted to arxiv on: 31 Jan 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); 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 This paper proposes a novel framework for efficient model compression called Early Pruning with Self-Distillation (EPSD). The authors build upon recent work on knowledge distillation and network pruning to develop a two-step process that combines early pruning and self-distillation. EPSD identifies and preserves distillable weights in the pruned network, enabling it to favor self-distillation and improve training for compression. The framework is evaluated on various benchmarks, including CIFAR-10/100, Tiny-ImageNet, full ImageNet, CUB-200-2011, and Pascal VOC, demonstrating improved performance compared to advanced pruning and self-distillation techniques. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper creates a new way to make AI models smaller and faster. It combines two ideas: getting rid of unnecessary parts in the model (early pruning) and helping the model learn from itself (self-distillation). The combination of these two ideas, called EPSD, makes it easier for the model to learn and improves its performance on various tasks. The authors tested EPSD on different datasets and found that it worked better than other methods. |
Keywords
* Artificial intelligence * Distillation * Knowledge distillation * Model compression * Pruning
