Boltzmann Sampling by Diabatic Quantum Annealing
AI Breakdown
Get a structured breakdown of this paper — what it's about, the core idea, and key takeaways for the field.
Abstract
Boltzmann sampling plays a key role in numerous algorithms, including those in machine learning. While quantum annealers have been explored as fast Boltzmann samplers, their reliance on environmental noise limits control over the effective temperature, introducing uncertainties in the sampling process. As an alternative, we propose diabatic quantum annealing -- a faster, purely unitary process -- as a controllable Boltzmann sampler, where the effective temperature is tuned via the annealing rate. Using infinite-range and two-dimensional ferromagnetic Ising models, we show that this approach enables rapid and accurate sampling in the high-temperature regime, with errors remaining bounded in the paramagnetic phase, regardless of system size.