Towards application of machine learning in quantum state estimation for photonic systems

Abstract

This project compare two techniques used for quantum state tomography: the traditional technique of Maximum Likelihood Estimation(MLE), and recent technique based on Conditional Generative Adversarial Networks(CGAN). MLE was used to analyze single and double qubit data and to retrieve the density matrix. Then we constructed the CGAN model using Deep neural networks. The perfomance of CGAN was tested against MLE for single qubit tomography. Then the CGAN model was trained on different data sets to see if the perfomance changed. The CGAN model was constantly compared with the MLE to see for perfomance. Then the CGAN architecture was applied to a two-qubit quantum state. MLE techinique performed superior to CGAN in both single and double qubit state tomography. CGAN was unable to produce states with complex entries in thier density matrix.