Design of Optical Logic Gates Using Mach–Zehnder Interferometers and Machine Learning

In the present work, a technique for designing all-optical logic gates from a Mach-Zehnder Interferometer (MZI) using machine learning (ML) is presented. The main idea behind the proposed approach is to tune the phase shifts of the MZI in such a way that it works a desired logic gate. The phase shifts are estimated as the ones that provide the higher distance between the outputs corresponding to the logic levels 0 and the 1. Moreover, instead of using a full scan of phase shifts, as in previous works, the proposed technique may use databases with a much smaller amount of data, using a ML technique to estimate the other outputs. In particular, four regression methods are tested (one at a time) for fitting the output of the MZI. Simulation results that illustrate and evaluate several aspects of the proposed techniques are presented.