Quantum Amplitude Amplification for Reinforcement Learning
Reinforcement learning models require a choice rule for assigning probabilities to actions during learning. This chapter reviews work on a new choice rule based on an amplitude amplification algorithm originally developed in quantum computing. The basic theoretical ideas for amplitude amplification are reviewed, as well as four different simulation applications to a task with a predator learning to catch a prey in a grid world, and one application to human learning in a 4 arm bandit choice task. The applications reviewed in this chapter demonstrate that QRL can improve speed and robustness of learning compared to standard choice rules such as epsilon-greedy and softmax choice rules.
K. Rajagopal et al., "Quantum Amplitude Amplification for Reinforcement Learning," Handbook of Reinforcement Learning and Control, Studies in Systems, Decision and Control, vol. 325, pp. 819-833, Springer, Jun 2021.
The definitive version is available at https://doi.org/10.1007/978-3-030-60990-0_26
Mechanical and Aerospace Engineering
International Standard Serial Number (ISSN)
Book - Chapter
© 2021 Springer, All rights reserved.
24 Jun 2021