Regensburg 2022 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
QI: Fachverband Quanteninformation
QI 6: Quantum Information: Concepts and Methods
QI 6.12: Talk
Tuesday, September 6, 2022, 12:45–13:00, H9
Fiber communication with collective quantum measurements: a machine learning perspective with applications — •Matteo Rosati1 and Janis Nötzel2 — 1Electrical Engineering and Computer Science, Technische Universität Berlin, 10587 Berlin, Germany — 21Emmy-Noether Gruppe Theoretisches Quantensystemdesign Lehrstuhl für Theoretische Informationstechnik Technische Universität München.
The transmission rate of classical bits on optical fiber is ultimately governed by the Holevo capacity. Achieving such rate requires writing information into coherent states of light and then performing a collective quantum measurement on multiple received signals at once, known as quantum joint-detection receiver (QJDR).
We find that the realization of a QJDR would enable two key advantages in current communication networks: (i) an estimated 55% decrease in energy consumption of optical amplifiers; (ii) an unbounded logarithmic growth of the channel capacity with the signal pulse rate, as opposed to the bounded rate attained by conventional detectors.
We then develop a machine learning framework to discover approximate implementations of the QJDR with a state-of-the-art photonic circuit. We compute the theoretical learning complexity of such photonic circuits, showing that it is polynomial in the number of optical modes, and introduce a simple algorithm to optimize them. Finally, we show that our algorithm is able to discover decoder setups that are both realizable at the state of the art and can attain a decoding success rate as high as 93% of the optimal QJDR.