Summary of Scaling Value Iteration Networks to 5000 Layers For Extreme Long-term Planning, by Yuhui Wang et al.
Scaling Value Iteration Networks to 5000 Layers for Extreme Long-Term Planning
by Yuhui Wang, Qingyuan Wu, Weida Li, Dylan R. Ashley, Francesco Faccio, Chao Huang, Jürgen Schmidhuber
First submitted to arxiv on: 12 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 Value Iteration Network (VIN) is an end-to-end differentiable architecture for planning in reinforcement learning (RL). However, VINs struggle to scale to long-term and large-scale planning tasks. To address this, the authors augment the latent MDP with a dynamic transition kernel, improving its representational capacity, and introduce an “adaptive highway loss” to mitigate the vanishing gradient problem. The proposed method, Dynamic Transition VIN (DT-VIN), is evaluated on 2D maze navigation environments and the ViZDoom 3D navigation benchmark, demonstrating improved performance and scalability. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The Value Iteration Network (VIN) is a new way for computers to plan ahead in complex situations. Right now, it’s hard for computers to make good decisions over a long period of time or when there are many possible paths to choose from. To fix this, the researchers came up with two ideas: making the computer’s internal map better and helping the planning process by making connections between different steps. This new method, called Dynamic Transition VIN (DT-VIN), is able to plan ahead much more effectively and solve complex problems. |
Keywords
» Artificial intelligence » Reinforcement learning
