Summary of Safe Reinforcement Learning in Black-box Environments Via Adaptive Shielding, by Daniel Bethell et al.
Safe Reinforcement Learning in Black-Box Environments via Adaptive Shielding
by Daniel Bethell, Simos Gerasimou, Radu Calinescu, Calum Imrie
First submitted to arxiv on: 28 May 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: Machine Learning (cs.LG)
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 In this paper, researchers propose a new approach called ADVICE (Adaptive Shielding with a Contrastive Autoencoder) for safely exploring reinforcement learning (RL) agents in unknown environments. The goal is to reduce the risk of executing harmful actions during training. The authors introduce a post-shielding technique that distinguishes safe and unsafe features of state-action pairs, allowing the agent to learn from its experiences while avoiding dangerous outcomes. The results show that ADVICE significantly outperforms existing methods in reducing safety violations during training, with comparable outcome rewards. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine you’re teaching an AI how to play a game or make decisions without knowing the rules beforehand. This can be risky because the AI might do something bad if it doesn’t understand the consequences. To solve this problem, scientists created a new way called ADVICE that helps train AI agents safely in unknown environments. It does this by identifying safe and unsafe actions during training, so the agent learns from its experiences without causing harm. |
Keywords
» Artificial intelligence » Autoencoder » Reinforcement learning
