Summary of Scaling Laws For the Value Of Individual Data Points in Machine Learning, by Ian Covert et al.
Scaling Laws for the Value of Individual Data Points in Machine Learning
by Ian Covert, Wenlong Ji, Tatsunori Hashimoto, James Zou
First submitted to arxiv on: 30 May 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 introduces a novel perspective on machine learning model training by investigating the relationship between individual data points’ contributions to model performance and the size of the dataset. The authors find that this contribution shrinks predictably with the dataset’s size in a log-linear manner, but with significant variability among different data points. This variability suggests that certain points are more valuable in small datasets while others are more useful in large datasets. The paper proposes learning theory to support these findings and demonstrates applications of individualized scaling laws to data valuation and subset selection. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The researchers looked at how machine learning models improve when you give them more training data. They found that each piece of data helps a little bit, but the amount it helps shrinks as the dataset gets bigger. This is true for different types of models, too! The team also came up with ways to measure and use this information to make decisions about which data points are most important. |
Keywords
» Artificial intelligence » Machine learning » Scaling laws
