Summary of Learning Iterative Reasoning Through Energy Diffusion, by Yilun Du et al.
Learning Iterative Reasoning through Energy Diffusion
by Yilun Du, Jiayuan Mao, Joshua B. Tenenbaum
First submitted to arxiv on: 17 Jun 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
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 We introduce Iterative Reasoning through Energy Diffusion (IRED), a novel framework for learning to reason and make decisions. IRED formulates problems with energy-based optimization, learns energy functions to represent constraints between input conditions and desired outputs. This framework adapts the number of optimization steps during inference based on problem difficulty, enabling it to solve complex problems outside its training distribution. Key techniques include annealed energy landscapes for easier inference and score function and energy landscape supervision for faster and more stable training. Our experiments show that IRED outperforms existing methods in continuous-space reasoning, discrete-space reasoning, and planning tasks, particularly in challenging scenarios. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper introduces a new way to learn and make decisions called Iterative Reasoning through Energy Diffusion (IRED). It uses energy-based optimization to solve problems and adapt to new situations. The framework is good at solving complex puzzles like Sudoku, completing matrices with large values, and finding paths in graphs. The authors use two special techniques to make the method work better: a sequence of easier-to-solve energy landscapes and supervision during training. Their tests show that IRED is better than other methods at making decisions. |
Keywords
* Artificial intelligence * Diffusion * Inference * Optimization
