Summary of Hybrid Recurrent Models Support Emergent Descriptions For Hierarchical Planning and Control, by Poppy Collis et al.
Hybrid Recurrent Models Support Emergent Descriptions for Hierarchical Planning and Control
by Poppy Collis, Ryan Singh, Paul F Kinghorn, Christopher L Buckley
First submitted to arxiv on: 20 Aug 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: Systems and Control (eess.SY)
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 A novel hybrid state-space model, recurrent switching linear dynamical systems (rSLDS), is proposed to learn discrete abstractions for solving continuous problems. Building upon previous work, this approach discovers meaningful behavioural units via piecewise linear decomposition and models how underlying continuous states drive discrete mode switches. The rich representations formed by rSLDS can provide useful abstractions for planning and control. A hierarchical model-based algorithm inspired by Active Inference is presented, comprising a discrete MDP sitting above a low-level linear-quadratic controller. This approach enables specification of temporally-abstracted sub-goals, lifting exploration into discrete space to exploit information-theoretic bonuses, and caching approximate solutions to low-level problems in the discrete planner. The proposed model is applied to the Continuous Mountain Car task, demonstrating fast system identification via enhanced exploration and non-trivial planning through abstract sub-goal delineation. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A team of researchers has created a new way for computers to learn about complex things by breaking them down into smaller, more manageable pieces. This helps computers make better decisions and plans for tasks that involve lots of variables and options. The approach uses something called “recurrent switching linear dynamical systems” (rSLDS) to figure out what’s happening in a situation and then use that information to decide what to do next. This can help computers learn more efficiently and make better choices. |
Keywords
* Artificial intelligence * Inference
