Summary of Group Benefits Instances Selection For Data Purification, by Zhenhuang Cai et al.
Group Benefits Instances Selection for Data Purification
by Zhenhuang Cai, Chuanyi Zhang, Dan Huang, Yuanbo Chen, Xiuyun Guan, Yazhou Yao
First submitted to arxiv on: 23 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Multimedia (cs.MM)
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 A deep learning model training method, named GRIP, is proposed to alleviate the problem of label noise in web-based training data. The approach combines group regularization with instance purification to reduce overfitting and improve classification performance on both synthetic and real-world noisy datasets. Soft label supervision is used to learn inter-class similarities, while an instance purification operation identifies noisy labels by measuring their difference from class soft labels. The proposed method outperforms existing state-of-the-art methods in comprehensive experimental results. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A new way to train deep learning models uses web images instead of manually annotated datasets. This makes training faster and more efficient. However, the web data often has mistakes or “noisy” labels that can hurt the model’s performance. Existing solutions for fixing noisy labels don’t work well on real-world data. The GRIP method is designed to solve this problem by using a special type of regularization and an operation to clean up noisy labels. This approach combines the benefits of different methods and improves the model’s performance. |
Keywords
* Artificial intelligence * Classification * Deep learning * Overfitting * Regularization
