Physics-inspired Generative AI models via real hardware-based noisy quantum diffusion
The research proposes physics-inspired Quantum Diffusion Models (QDMs) to improve image generation quality for MNIST. They tested a hybrid quantum-classical stochastic walk dynamics, showing it produces statistically more robust models with lower Fréchet Inception Distance (FID) than purely classical methods. Furthermore, they demonstrated an algorithm that exploits the intrinsic noise of real IBM quantum hardware to generate images.