Genetic algorithms to learn an unknown linear transformation

Integrated photonics enabled the capability to implement complex interferometers with a large number of modes. The applications of this experimental toolbox can be found in different quantum information tasks, including quantum walks and simulation [1], quantum metrology [2], up to the implementation of Boson Sampling [3-6], one of the most promising candidates to reach experimentally the regime where a quantum device solves a computational problem faster than its classical counterpart. Both from a fundamental aspect and for characterization purposes, it becomes crucial to identify tools allowing to learn the action of an unknown transformation