Докладчик: Илья Лучников (МФТИ, Сколтех)
Дата: 21 июля 2020
Семинар проходит в режиме видеоконференции. Трансляция семинара будет проводиться в режиме реального времени на YouTube-канале ЦКТ
Many data processing tasks in quantum technologies can be reformulated as optimization problems. Some optimization problems include natural "quantum" constraints on the parameters of a problem. For example, to perform tomography of a quantum state, one needs to find a density matrix maximizing a likelihood function and guarantee the positivity of the density matrix and that its trace is equal to one. In many cases, constraints form a Riemannian manifold (smooth multidimensional surface), which allows turning from constrained optimization in the Euclidean space to unconstrained optimization on a Riemannian manifold. We apply this idea to different optimization tasks in quantum technologies. We also provide QGOpt library, written on top of TensorFlow, that allows performing Riemannian optimization on the most common "quantum" manifolds. During the talk, I will show different use cases of the Riemannian optimization in quantum technologies and present our open source library QGOpt.