25.03.2026 Adrián Pérez-Salinas (ETH Zurich)
Location: HIT E 41.1, Time: 12:00
The subtleties of sampling hardness in quantum systems
Sampling tasks have been successful in establishing quantum advantages both in theory and experiments. This has fueled the use of quantum computers for generative modeling to create samples following the probability distribution underlying a given dataset. In particular, the potential to build generative models on classically hard distributions would immediately preclude classical simulability, due to theoretical separations. In this talk, we will review the foundations of sampling hardness and investigate its consequences. We will visit quantum generative models from the perspective of output distributions, showing that models that anticoncentrate are not trainable on average, including those exhibiting quantum advantage. This observed trade-off is linked to verification of quantum processes. In addition, this phenomenon prevents the use of classical simulation for physical processes that require sampling from a large space for their computation (WIP). In contrast, quantum processes sparse output distributions and much more manageable, and still compatible with quantum-classical separations.