Summary of Asymptotic Unbiased Sample Sampling to Speed Up Sharpness-aware Minimization, by Jiaxin Deng et al.
Asymptotic Unbiased Sample Sampling to Speed Up Sharpness-Aware Minimization
by Jiaxin Deng, Junbiao Pang, Baochang Zhang
First submitted to arxiv on: 12 Jun 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
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 presents a novel approach called Asymptotic Unbiased Sampling with respect to iterations (AUSAM) that improves the efficiency of Sharpness-Aware Minimization (SAM) while maintaining its generalization capacity. AUSAM accelerates SAM by probabilistically sampling beneficial data points for optimization based on the Gradient Norm of each Sample (GNS). This approach achieves a speedup of over 70% compared to SAM, making it suitable for various tasks and networks, including classification, human pose estimation, and network quantization. AUSAM also outperforms recent dynamic data pruning methods in terms of maintaining performance. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps make machine learning faster and better. Researchers found a way to make an important technique called Sharpness-Aware Minimization (SAM) work more efficiently without sacrificing its ability to learn from data accurately. They did this by cleverly selecting which data points are most important for SAM to use, based on how much the algorithm is trying to improve. This new approach, called AUSAM, works well for different types of tasks and can even be used with other techniques to make machine learning even more powerful. |
Keywords
» Artificial intelligence » Classification » Generalization » Machine learning » Optimization » Pose estimation » Pruning » Quantization » Sam
