Summary of Learning Action-based Representations Using Invariance, by Max Rudolph et al.
Learning Action-based Representations Using Invariance
by Max Rudolph, Caleb Chuck, Kevin Black, Misha Lvovsky, Scott Niekum, Amy Zhang
First submitted to arxiv on: 25 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); 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 This paper presents a method for robust reinforcement learning agents to identify relevant state features from high-dimensional observations, even when faced with many distractions. The proposed approach, called action-bisimulation encoding, builds upon the concept of controllability and introduces a recursive invariance constraint to capture long-horizon elements that affect agent control. The authors demonstrate the effectiveness of this method by pretraining agents on reward-free data and showing improved sample efficiency in various environments, including a photorealistic 3D simulation domain. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about helping machines learn from their environment by figuring out what’s important to pay attention to. Right now, it’s hard for machines to understand what they need to do to control things that happen far away, like a wall approaching. The researchers created a new way to teach machines called action-bisimulation encoding. It helps machines learn what matters and what doesn’t, making them better at doing tasks in the long run. |
Keywords
* Artificial intelligence * Attention * Pretraining * Reinforcement learning
