A learning theory for quantum photonic processors and beyond
Seminar author:Matteo Rosati
Event date and time:12/13/2022 04:00:pm
Event location:
Event contact:
We consider a family of Gaussian and non-Gaussian continuous-variable (CV) circuits, suitable to describe state-of-the-art photonic processors, and evaluate its learning capabilities.
Our basic learning problem is: given copies of an unknown quantum state, we apply to each of them a random quantum circuit from a given set C and obtain measurement outcomes. The objective is to approximate the unknown state using a state from a given hypothesis set S, using a small number of samples. We show that a good approximation can be found with a number of samples polynomial in the number of modes of the CV circuit, which is a measure of its
size. We apply our results to learning a decoder for optical communication, outperforming the state of the art. Ref. https://arxiv.org/abs/2209.
Our basic learning problem is: given copies of an unknown quantum state, we apply to each of them a random quantum circuit from a given set C and obtain measurement outcomes. The objective is to approximate the unknown state using a state from a given hypothesis set S, using a small number of samples. We show that a good approximation can be found with a number of samples polynomial in the number of modes of the CV circuit, which is a measure of its
size. We apply our results to learning a decoder for optical communication, outperforming the state of the art. Ref. https://arxiv.org/abs/2209.