Summary of A Note on Loss Functions and Error Compounding in Model-based Reinforcement Learning, by Nan Jiang
A Note on Loss Functions and Error Compounding in Model-based Reinforcement Learning
by Nan Jiang
First submitted to arxiv on: 15 Apr 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 The paper sheds light on the theoretical understanding of model-based reinforcement learning (MBRL) in the context of deep reinforcement learning. MBRL has received a bad empirical reputation due to error compounding, but its superior theoretical properties are not fully utilized. The study reconciles this discrepancy and highlights the limitations of empirically popular losses, such as MuZero loss. Specifically, it shows that MuZero loss fails in stochastic environments and suffers from exponential sample complexity in deterministic environments when data provides sufficient coverage. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper is about understanding how to use computers to make good decisions by learning from experience. It talks about a way of doing this called model-based reinforcement learning. This method has been shown to be effective, but there are some problems with it. The study tries to solve these problems and understand why they happen. It also looks at the limitations of some popular ways of training models. |
Keywords
» Artificial intelligence » Reinforcement learning
