Summary of Criterion Collapse and Loss Distribution Control, by Matthew J. Holland
Criterion Collapse and Loss Distribution Control
by Matthew J. Holland
First submitted to arxiv on: 15 Feb 2024
Categories
- Main: Machine Learning (stat.ML)
- Secondary: Machine Learning (cs.LG)
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 delves into the phenomenon of “criterion collapse” in machine learning, where optimizing one metric leads to optimality in another. The authors focus on identifying conditions for collapse under various learning criteria, including DRO and OCE risks, CVaR, tilted ERM, Flooding, and SoftAD algorithms. They demonstrate that criterion collapse extends beyond existing results for CVaR and DRO, and further explore surrogate losses to show when monotonic criteria like tilted ERM cannot avoid collapse, whereas non-monotonic alternatives can. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about how machine learning models can get stuck in a cycle where they’re optimized for one thing, but end up doing something else. The researchers looked at different ways of measuring how well a model performs and found that sometimes these measurements can “collapse” into each other. They wanted to know when this happens and under what conditions. By studying things like DRO and OCE risks, CVaR, and tilted ERM, they found some surprising results about what happens when models are optimized for certain metrics. |
Keywords
* Artificial intelligence * Machine learning
