Abstract:Generative models realized with machine learning techniques are powerful tools to infer complex and unknown data distributions from a finite number of training samples in order to produce new synthetic data. Diffusion models are an emerging framework that have recently overcome the performance of the generative adversarial networks in creating synthetic text and high-quality images. Here, we propose and discuss the quantum generalization of diffusion models, i.e., three quantum-noise-driven generative diffusion models that could be experimentally tested on real quantum systems. The idea is to harness unique quantum features, in particular the non-trivial interplay among coherence, entanglement and noise that the currently available noisy quantum processors do unavoidably suffer from, in order to overcome the main computational burdens of classical diffusion models during inference. Hence, we suggest to exploit quantum noise not as an issue to be detected and solved but instead as a very remarkably beneficial key ingredient to generate much more complex probability distributions that would be difficult or even impossible to express classically, and from which a quantum processor might sample more efficiently than a classical one. An example of numerical simulations for an hybrid classical-quantum generative diffusion model is also included. Therefore, our results are expected to pave the way for new quantum-inspired or quantum-based generative diffusion algorithms addressing more powerfully classical tasks as data generation/prediction with widespread real-world applications ranging from climate forecasting to neuroscience, from traffic flow analysis to financial forecasting.

Macroscopic Noise Amplification by Asymmetric Dyads in Non-Hermitian Optical Systems for Generative Diffusion Models

Macroscopic noise amplification by asymmetric dyads in non-Hermitian optical systems for generative diffusion models

Anomalous transient amplification of waves in non-normal photonic media

Photonic diffractive generators through sampling noises from scattering media

Quantum-Noise-Driven Generative Diffusion Models

Overdispersion in gate tomography: Experiments and continuous, two-scale random walk model on the Bloch sphere

Scalable photonic diffractive generators through sampling noises from scattering medium

Complex-valued Retrievals From Noisy Images Using Diffusion Models

Optomechanical sideband asymmetry explained by stochastic electrodynamics

Noiseless Linear Amplification and Distillation of Entanglement

Optical Diffusion Models for Image Generation

Efficient noiseless linear amplification for light fields with larger amplitudes

Modeling of Fluctuations in Dynamical Optoelectronic Device Simulations within a Maxwell-Density Matrix Langevin Approach

Non-Hermitian dynamics and nonreciprocity of optically coupled nanoparticles

Symmetric and asymmetric discrimination of bosonic loss: Toy applications to biological samples and photo-degradable materials

Iso-Diffusion: Improving Diffusion Probabilistic Models Using the Isotropy of the Additive Gaussian Noise

Nonlinear optical pulses in media with asymmetric gain

Generalized Diffusion Model with Adjusted Offset Noise

Fitting quantum noise models to tomography data

Photonic probabilistic machine learning using quantum vacuum noise

Accounting for Inhomogeneous Broadening in Nano-Optics by Electromagnetic Modeling Based on Monte Carlo Methods