all AI news
Quantum Policy Iteration via Amplitude Estimation and Grover Search -- Towards Quantum Advantage for Reinforcement Learning. (arXiv:2206.04741v1 [quant-ph])
cs.LG updates on arXiv.org arxiv.org
We present a full implementation and simulation of a novel quantum
reinforcement learning (RL) method and mathematically prove a quantum
advantage. Our approach shows in detail how to combine amplitude estimation and
Grover search into a policy evaluation and improvement scheme. We first develop
quantum policy evaluation (QPE) which is quadratically more efficient compared
to an analogous classical Monte Carlo estimation and is based on a quantum
mechanical realization of a finite Markov decision process (MDP). Building on
QPE, we …
amplitude arxiv iteration learning policy quantum quantum advantage reinforcement reinforcement learning search