Summary of Explaining Bayesian Optimization by Shapley Values Facilitates Human-ai Collaboration, By Julian Rodemann et al.
Explaining Bayesian Optimization by Shapley Values Facilitates Human-AI Collaboration
by Julian Rodemann, Federico Croppi, Philipp Arens, Yusuf Sale, Julia Herbinger, Bernd Bischl, Eyke Hüllermeier, Thomas Augustin, Conor J. Walsh, Giuseppe Casalicchio
First submitted to arxiv on: 7 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC); Robotics (cs.RO); 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 optimization with Gaussian processes has become a crucial algorithm for black box optimization problems. However, it often lacks transparency in its decision-making process, making it challenging to provide reasons for the proposed parameters. This is particularly relevant in human-in-the-loop applications of BO, such as robotics. The proposed ShapleyBO framework addresses this issue by using game-theoretic Shapley values to quantify each parameter’s contribution to BO’s acquisition function. This enables identifying how strongly each parameter drives exploration and exploitation for additive acquisition functions like the confidence bound. Additionally, ShapleyBO can disentangle contributions to exploration into aleatoric and epistemic uncertainty. The method also gives rise to a ShapleyBO-assisted human-machine interface, allowing users to intervene when proposals do not align with their reasoning. This HMI demonstrates benefits for personalizing wearable robotic devices (assistive back exosuits) by human-in-the-loop BO. Results show that teams with access to ShapleyBO can achieve lower regret than those without. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary ShapleyBO is a new way to understand how Bayesian optimization makes decisions. Right now, it’s hard for humans and computers to work together because they don’t know why the computer is suggesting certain things. The authors of this paper created a tool called ShapleyBO that helps figure out which factors are most important in making those suggestions. This can be really helpful when working with robots or other machines that need human input. |
Keywords
* Artificial intelligence * Optimization
