AI and the World's Hard Problems · Quantum physics & quantum computing · arXiv · 2026-10-23
Quantum Fokker-Planck solvers as trainable generative models
The authors develop generative models built on quantum solvers for stochastic differential equations. Instead of layered variational circuits, they parameterize the potential of a Langevin process and train the model to match empirical distributions. The model recovers parameters of multivariate distributions and remains trainable at scale.