Summary of Slow and Steady Wins the Race: Maintaining Plasticity with Hare and Tortoise Networks, by Hojoon Lee et al.
Slow and Steady Wins the Race: Maintaining Plasticity with Hare and Tortoise Networks
by Hojoon Lee, Hyeonseo Cho, Hyunseung Kim, Donghu Kim, Dugki Min, Jaegul Choo, Clare Lyle
First submitted to arxiv on: 1 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 The study investigates the loss of generalization ability in neural networks, revisiting warm-starting experiments. It reveals that common methods designed to enhance plasticity provide limited benefits to generalization. The paper introduces the Hare & Tortoise method, consisting of two components: the Hare network and the Tortoise network. By periodically reinitializing the Hare network to the Tortoise’s weights, the method preserves plasticity while retaining general knowledge. This improves advanced reinforcement learning algorithms on the Atari-100k benchmark. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how neural networks can forget what they’ve learned, even when trying to learn new things. It shows that most ways to help them remember don’t actually work very well. The researchers came up with a new idea called Hare & Tortoise, which is like two parts working together. One part (Hare) learns quickly, and the other part (Tortoise) helps it remember what it’s learned before. This helps the network keep learning and remembering things at the same time. |
Keywords
» Artificial intelligence » Generalization » Reinforcement learning
