Quantum-tailored machine-learning characterization of a superconducting qubit

Abstract

Machine learning (ML) is a promising approach for performing challenging quantum-information tasks such as device characterization, calibration and control. ML models can train directly on the data produced by a quantum device while remaining agnostic to the quantum nature of the learning task. However, these generic models lack physical interpretability and usually require large datasets in order to learn accurately. Here we incorporate features of quantum mechanics in the design of our ML approach to characterize the dynamics of a quantum device and learn device parameters. This physics-inspired approach outperforms physics-agnostic recurrent neural networks trained on numerically generated and experimental data obtained from continuous weak measurement of a driven superconducting transmon qubit. This demonstration shows how leveraging domain knowledge improves the accuracy and efficiency of this characterization task, thus laying the groundwork for more scalable characterization techniques.

Monitoring fast superconducting qubit dynamics using a neural network

“…To further improve the LSTM accuracy for applications such as parameter estimation, it is possible to add physical constraints to the LSTM loss function [47,50,51]. In addition, other neural network architectures such as tensor networks or models based on dilated causal convolutions [52] may improve prediction accuracy, though such comparisons require more training data.…”

“…3 and 4 demonstrate accurate estimation of various decay rates, measurement efficiency and the memory time of the system from a simple trajectory decomposition that requires no prior knowledge of the resonator memory. This makes the LSTM a useful tool in the context of parameter estimation from weak measurements [46,47].…”

“…During the training weights and biases of the network are adjusted to minimize a cost function L, which contains the following three components (similar to Ref. [47]): • A cross-entropy loss associated with the projective measurement result at the end of a voltage record…”

“…To further assess the LSTM performance, more trajectories are needed to reduce our uncertainty in the estimate of the true quantum state. Alternatively, it is possible to train the LSTM on simulated trajectories for which the true quantum state is known [47].…”