Manipulation of Spin Dynamics by Deep Reinforcement Learning Agent

TL;DR

This paper demonstrates how a deep reinforcement learning agent can effectively control spin dynamics in quantum systems, outperforming traditional methods and providing interpretable policies for preparing spin squeezed states.

Contribution

It introduces a reinforcement learning approach using PPO for quantum spin control, achieving superior policies in many-body systems and offering insights into quantum dynamics.

Findings

RL policies outperform greedy and adiabatic methods

RL provides physically interpretable control strategies

Effective in both mean-field and many-body quantum systems

Abstract

We implement the reinforcement learning agent for a spin-1 atomic system to prepare spin squeezed state from given initial state. Proximal policy gradient (PPO) algorithm is used to deal with continuous external control field and final optimized protocol is given by a stochastic policy. In both mean-field system and two-body quantum system, RL agent finds the optimal policies. In many-body quantum system, it also gives polices that outperform purely greedy policy and optimized adiabatic passage. These polices given by RL agent have good physical interpretability in phase space and may help us to understand quantum dynamics. In fact, RL could be highly versatile in quantum optimal control problems.