Summary of 2d-oob: Attributing Data Contribution Through Joint Valuation Framework, by Yifan Sun et al.
2D-OOB: Attributing Data Contribution Through Joint Valuation Framework
by Yifan Sun, Jingyan Shen, Yongchan Kwon
First submitted to arxiv on: 7 Aug 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
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 proposed 2D-OOB framework addresses limitations in existing data valuation methods by jointly determining helpful samples and noisy cells. Building on the concept of data valuation, this paper aims to recognize varying quality within individual data points. A single scalar score cannot accurately capture this distinction, as shown through abnormal data points with both clean and noisy cells. The authors introduce 2D-OOB, an out-of-bag estimation framework that achieves state-of-the-art performance in multiple use cases while being exponentially faster than existing methods. The results demonstrate promising outcomes in detecting fine-grained outliers at the cell level and localizing backdoor triggers in data poisoning attacks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A new way of measuring the importance of individual data points has been developed, called 2D-OOB. Right now, this type of evaluation only gives a single score for each piece of data, which doesn’t take into account that some parts of the data might be good while others are bad. The authors created a better system to figure out which data points are helpful or hurtful and which parts of those points are causing problems. This new method is really fast and does a great job of finding tiny mistakes in the data, like when someone tries to trick a machine learning model by adding fake information. |
Keywords
» Artificial intelligence » Machine learning
