Reproduce “Model-based Offline Quantum Reinforcement Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Simon Eisenmann, Daniel Hein, Steffen Udluft, Thomas A. Runkler
Researchers from Technical University of Munich and Siemens AG introduce the first algorithm for model-based offline quantum reinforcement learning. The approach utilizes variational quantum circuits as both an environment's surrogate model and the policy, successfully learning to balance a cart-pole from a static dataset and achieving "perfect policies" in seven out of ten optimization runs.
Key insights
- Variational Quantum Circuits can effectively learn and model the dynamics of control environments from offline data, serving as viable surrogate models for policy optimization.
- Model-based offline QRL provides a feasible and robust framework for training quantum agents from static datasets, mitigating the high costs and interaction challenges of online quantum learning.
- The data re-uploading technique significantly enhances the predictive power and expressive capabilities of VQC surrogate models, improving accuracy by an order of magnitude with just one re-upload.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.