Machine Learning Assisted Bayesian Convex Decomposition for Mixed Quantum States

HsienYi Hsieh^1*, Ray-Kuang Lee¹

¹Institute of Photonics Technologies, National Tsing Hua University, Hsinchu, Taiwan

* Presenter:HsienYi Hsieh, email:moro1905@hotmail.com

We analyze experimentally generated non-Gaussian optical quantum states using a
flow-based machine learning framework. The normalizing flow model performs Bayesian inference of physical parameters, incorporating prior knowledge to enhance reliability and quantify
uncertainties. This approach further enables a Bayesian decomposition of the quantum state,
revealing how pure states evolve into mixed ones under experimental imperfections. Our results
demonstrate that flow-based models provide an accurate and interpretable tool for characterizing
non-Gaussian quantum states

Keywords: Machine Learning, Normalizing Flow, Quantum Optics, Non-Gaussian States, Quantum Tomography