Machine learning assisted GaAsN circular polarimeter

Abstract We demonstrate the application of a two stage machine learning algorithm that enables one to correlate electrical signals from GaAs x N 1 − x circular polarimeter with intensity, degree polarization (DCP) and handedness an incident light beam. Specifically, we employ multimodal logistic regression discriminate six-layer neural network (NN) establish relationship between input voltages, intensity DCP. have developed particular NN training strategy substantially improves accuracy device. The was trained tested on theoretically generated photoconductivity photoluminescence experimental results. Even for small dataset (70 instances), it is shown proposed correctly predicts linear, right left circularly polarized light, misclassifying less than $1.5\%$?> 1.5 mathvariant="normal">% cases attaining larger $97\%$?> 97 in vast majority predictions ( $92\%$?> 92 ) These numbers are significantly improved datasets (4851 instances). versatile enough can be easily adjusted other device configurations where map needs established parameters response. Training testing data files as well provided supplementary material.