Summary of Mixed Dynamics in Linear Networks: Unifying the Lazy and Active Regimes, by Zhenfeng Tu et al.
Mixed Dynamics In Linear Networks: Unifying the Lazy and Active Regimes
by Zhenfeng Tu, Santiago Aranguri, Arthur Jacot
First submitted to arxiv on: 27 May 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 proposed unifying formula captures the training dynamics of linear networks across three regimes: lazy, balanced/active, and a mixed regime that combines elements of both. The mixed regime allows for rapid convergence from any random initialization, while avoiding the low-rank bias present in lazy dynamics. This framework provides an almost complete phase diagram of training behavior, shedding light on the relationships between initialization variance, network width, and MSE training task. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper studies how neural networks learn new skills. It shows that there are three ways for a network to train: it can be slow-moving (lazy), fast-moving (balanced/active), or somewhere in between. The middle option is special because it allows the network to quickly adapt from any random start, while still avoiding some limitations of slow-moving training. This discovery helps create a complete map of how different settings affect the learning process. |
Keywords
» Artificial intelligence » Mse
