Data for Deep Reinforcement Learning for Individual Atomic Control and Cooling

This record contains the raw and curated data, trained reinforcement-learning controller models, simulated trajectory caches, and data-extraction scripts supporting the article "Deep Reinforcement Learning for Individual Atomic Control and Cooling." The dataset contains 11,051 raw experimental HDF5 shots acquired using a single cesium atom confined in an optical tweezer and coupled to a high-finesse optical cavity. These data include feedback and no-feedback cooling measurements, dedicated post-cooling temperature measurements, probe-detuning and photon-count parameter scans, and experimental reinforcement-learning training records. The record also contains curated figure and table source data, trained controller models, simulated turning-point data from 40,000-episode evaluations of the reinforcement-learning and differentiator controllers, and scripts that regenerate the curated datasets from the raw HDF5 files. Together with the GitHub repository, these materials support reproduction of all data in figures and Table I in the associated article. Detailed file descriptions, software requirements, extraction instructions, and figure mappings are provided in README.md. Companion code repository: https://github.com/math2peters/deep_RL_cavity_feedback_cooling Associated preprint: https://arxiv.org/abs/2606.30765 Contact: Matthew L. Peters, [e-mail]