This presentation investigates the potential of applying reinforcement learning to quantum control settings through the theoretical framework of Markov decision processes (MDPs). Barry et al....Show moreThis presentation investigates the potential of applying reinforcement learning to quantum control settings through the theoretical framework of Markov decision processes (MDPs). Barry et al. formulated Quantum Observable MDPs (QOMDPs) as a model for quantum environments which Tamon claims to generalize with their introduced Quantum Partially Observable MDPs (QPOMDPs). We construct a formalism of behavioural equivalence of decision process models in order to evaluate expressibility of models through distinguishibility of models. We show that all quantum experiments can be described as POMDPs and specific environments can be modelled by varying types and formulations of decision processes with their respective advantages and disadvantages. By conducting experiments on a quantum cartpole environment, this research investigates the effects of varying environmental specifications on learning behavior and performance in quantum control problems generalized by QOMDPs in order to determine which setting is more appropriate for accurately modeling the dynamics of the system. The insights gained in this thesis can aid with appropriate model specification which is important for learning in quantum control settings. This research also contributes to the understanding of the practical implications of environmental specifications in quantum control problems, with findings having implications for the development of more effective and efficient learning algorithms tailored to quantum control settings.Show less
