"Reinforcement Learning (RL) is a machine learning paradigm based on the notion of agents taking actions within their environment, with the aim of maximising a defined reward. In this seminar, we will further introduce RL and how it is already being used in sample-efficient control algorithms at CERN to optimise various aspects related to accelerator performance.
Building on this foundation, we will go on to discuss free energy-based reinforcement learning (FERL) with clamped quantum Boltzmann machines (QBM) — a method that was shown to improve the learning efficiency by orders of magnitude compared to classical Q-learning algorithms. This work also extends the FERL approach to multi-dimensional optimisation problems and eliminates the restriction to discrete action-space environments, opening doors for a broader range of real-world applications.
We will discuss the results obtained with quantum annealing, employing both a simulator, and D-Wave quantum annealing hardware, as well as a comparison to classical RL methods. We will cover how the algorithms are evaluated for control problems at CERN, such as the AWAKE electron beam line, and for classical RL benchmarks of varying degree of complexity.”