Estimating unknown qubit phase under telegraph noises using recurrent neural network

We consider a system of qubits affected by an environmental random-telegraph process (RTP) noise and investigate the recently proposed idea of using a spectator qubit (SQ) as a noise probe to remove noise effects from a data qubit (DQ). By measuring the SQ at various times with particular measurement bases, their measurement readouts can be used to extract the information of the noise effects. In this work, we propose the use of recurrent neural network (RNN), with the Long short-term memory (LSTM) layers, to process the SQ's readouts and train the model to estimate the errors occurred in the DQ. We present numerical results of the DQ's and SQ's phases and show that the LSTM can estimate the unknown DQ's phase with the estimation accuracy depending on the SQ's noise sensitivity.