Summary of Is Data Valuation Learnable and Interpretable?, by Ou Wu et al.
Is Data Valuation Learnable and Interpretable?
by Ou Wu, Weiyao Zhu, Mengyang Li
First submitted to arxiv on: 3 Jun 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 This paper investigates the value of individual samples in various data-driven tasks, such as training deep learning models. The authors build upon recent developments in data valuation methods, primarily based on the Shapley value from game theory. However, current approaches lack interpretability, which is crucial for applications like data pricing. To address this limitation, the researchers propose two novel frameworks: one using multi-layer perceptions (MLPs) and another employing regression trees as base models. These frameworks are designed to provide learned valuation models with fixed parameters, reusability, and interpretability, allowing samples to be explained as valuable or invaluable. Experimental results on benchmark datasets confirm that data valuation can be both learnable and interpretable. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps us understand how important individual pieces of data are in many computer-related tasks. Right now, people are working hard to figure out ways to value these data points. One popular approach is based on an idea from game theory called the Shapley value. But this method has some limitations, especially when it comes to explaining why certain data points are valuable or not. The researchers in this study wanted to see if they could create a new way to value data that’s both accurate and easy to understand. They came up with two new methods using special types of computer models. These models can help us learn how to value data in a more reliable and transparent way. |
Keywords
» Artificial intelligence » Deep learning » Regression
