Summary of Numeric Reward Machines, by Kristina Levina et al.
Numeric Reward Machines
by Kristina Levina, Nikolaos Pappas, Athanasios Karapantelakis, Aneta Vulgarakis Feljan, Jendrik Seipp
First submitted to arxiv on: 30 Apr 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- 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 The paper presents a novel extension to reward machines, which are typically used in reinforcement learning to guide agents towards optimal behavior. Reward machines currently rely on Boolean features, limiting their application to inherently numeric tasks. The authors propose two new types of reward machines: numeric-Boolean and numeric. The former emulates numeric features using Boolean combinations, while the latter uses the original numeric values alongside Boolean features. Experimental results in the Craft domain show that these approaches outperform a baseline reward machine when using cross-product Q-learning, Q-learning with counter-factual experiences, or the options framework for learning. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Reward machines help robots learn faster by telling them what’s good and bad. Right now, they only understand “yes” or “no” answers. What about tasks that need numbers? The researchers created new types of reward machines that can handle both “yes” or “no” answers and actual numbers. They tested these new approaches in a game where robots had to get close to a target. The results showed that their methods were much better than the old way. |
Keywords
» Artificial intelligence » Reinforcement learning
