Learning Quantum Data Distribution via Chaotic Quantum Diffusion Model

Generative models are pivotal for synthesizing data by modeling underlying distributions, with applications dominating fields like image and text generation. Quantum data, derived from quantum systems, introduce unique challenges that necessitate innovative approaches. We propose the chaotic quantum diffusion model, a framework that replaces the high-fidelity random circuits required by the quantum denoising diffusion probabilistic model with a more flexible, hardware-compatible process, leveraging the projected ensemble from chaotic time evolution dynamics. Our method achieves accuracy comparable to previous approaches while offering significant implementation efficiency. This approach broadens the applicability of quantum generative models across diverse quantum platforms.