Summary of Learning Linear Utility Functions From Pairwise Comparison Queries, by Luise Ge et al.
Learning Linear Utility Functions From Pairwise Comparison Queries
by Luise Ge, Brendan Juba, Yevgeniy Vorobeychik
First submitted to arxiv on: 4 May 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Machine Learning (stat.ML)
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 explores the learnability of linear utility functions from pairwise comparison queries in both passive and active learning settings. In the passive setting, it is efficiently possible to predict out-of-sample responses and approximately recover the true parameters of the utility function without noise or with Tsybakov noise when distributions are “nice”. However, without strong modeling assumptions, utility parameters cannot be learned for a large set of data distributions even with noise-free queries. In contrast, in an active learning setting, algorithms can efficiently learn both objectives and select pairwise queries to optimize utility learning. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper studies how well we can understand people’s preferences by asking them to compare things. We found that when we ask questions and get answers without any mistakes, we can easily predict what someone will prefer in the future and even figure out what makes something valuable or not. But if our data is mixed up with errors, it gets much harder to learn from it. When we let people choose which questions to answer, we found that we can still understand their preferences quite well. |
Keywords
» Artificial intelligence » Active learning
