Summary of Variational Entropy Search For Adjusting Expected Improvement, by Nuojin Cheng and Stephen Becker
Variational Entropy Search for Adjusting Expected Improvement
by Nuojin Cheng, Stephen Becker
First submitted to arxiv on: 17 Feb 2024
Categories
- Main: Machine Learning (stat.ML)
- Secondary: Machine Learning (cs.LG); Optimization and Control (math.OC)
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 As machine learning educators often employ Bayesian optimization to optimize black-box functions, we explore the connection between Expected Improvement (EI) and max-value entropy search (MES). Specifically, we show that EI can be viewed as a special case of MES when using variational inference (VI). Our work develops Variational Entropy Search (VES) methodology and VES-Gamma algorithm, which adapts EI by incorporating information-theoretic concepts. We demonstrate the effectiveness of VES-Gamma on various test functions and datasets, highlighting its theoretical and practical applications in Bayesian optimization. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Bayesian optimization is a way to find the best solution for a problem without knowing how it works. One popular method called Expected Improvement (EI) helps us choose the next step. Our research shows that EI can be seen as a special version of another method, max-value entropy search (MES), when we use something called variational inference (VI). We created new ways to do this, called Variational Entropy Search (VES) and VES-Gamma algorithm, which makes EI better by using ideas from information theory. We tested these methods on different problems and showed that they work well in situations where we want to find the best solution. |
Keywords
* Artificial intelligence * Inference * Machine learning * Optimization
