Summary of Learning Cognitive Maps From Transformer Representations For Efficient Planning in Partially Observed Environments, by Antoine Dedieu et al.
Learning Cognitive Maps from Transformer Representations for Efficient Planning in Partially Observed Environments
by Antoine Dedieu, Wolfgang Lehrach, Guangyao Zhou, Dileep George, Miguel Lázaro-Gredilla
First submitted to arxiv on: 11 Jan 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 paper introduces a novel transformer-based model called TDB that can learn an explicit world model of its environment and be used for planning or navigation. Unlike vanilla transformers, TDB’s latent codes compress the history of observations and actions, allowing it to predict future observations and extract interpretable cognitive maps of the environment. These maps are then paired with an external solver to solve constrained path planning problems. The paper shows that TDB trained on partially observed environments retains the predictive performance of vanilla transformers while solving shortest path problems exponentially faster. Additionally, TDB extracts interpretable representations from text datasets and reaches higher in-context accuracy than vanilla sequence models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper creates a special kind of AI model that can learn about its environment and use that knowledge to make decisions. The new model is called TDB and it’s really good at predicting what will happen next based on what has happened before. It’s like having a superpower! The researchers tested their model and found that it was much better than other models at solving problems, especially ones where you have to find the shortest path between two points. They also showed that TDB can be used with text data, which is really useful for things like chatbots or language translation. |
Keywords
* Artificial intelligence * Transformer * Translation
