Summary of Time-varying Gaussian Process Bandits with Unknown Prior, by Juliusz Ziomek et al.
Time-Varying Gaussian Process Bandits with Unknown Prior
by Juliusz Ziomek, Masaki Adachi, Michael A. Osborne
First submitted to arxiv on: 2 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: 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 Bayesian optimisation requires fitting a Gaussian process model, which in turn requires specifying prior on the unknown black-box function. However, most literature assumes this prior is known, but in reality, there are multiple possible priors for a given black-box function. To address this gap, we propose PE-GP-UCB, an algorithm capable of solving time-varying Bayesian optimisation problems without exact knowledge of the function’s prior. Our approach relies on rejecting wrong priors based on observed function values or accepting all candidate priors when consistent with observations. We provide a regret bound and empirically evaluate our algorithm on toy and real-world problems, showing it outperforms existing methods. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary PE-GP-UCB is a new way to solve problems where we don’t know the exact rules of a game or how something works. Usually, these problems need to fit a special kind of mathematical model called a Gaussian process. But sometimes, there are many possible models that could work, and we don’t know which one is correct. The algorithm PE-GP-UCB helps solve this problem by looking at the results we get when trying out different models. It can even handle problems that change over time! We tested it on some simple examples and real-world problems, and it did better than other methods. |
