Speaker
Description
Superconducting qubits are highly sensitive to their local environment. This sensitivity enables their use in quantum sensing, but it also makes them more susceptible to environmental noise. Background noise, device defects such as two-level systems, and changes in the qubit operating environment can all influence device performance. As a result, models developed under one set of conditions may not directly generalize to another. This work investigates Normalizing Flows, a class of machine-learning models that learn the probability distribution of a dataset, and how well they can model distributions across different superconducting qubit measurements. In this talk, their application as an anomaly-detection tool is evaluated across multiple qubit measurement modalities, where the learned probability distributions are used to identify deviations from expected device behavior. Building on these results, the approach is being extended to study how models trained in one qubit domain can be adapted to different measurements and experimental conditions, including whether limited fine-tuning can improve adaptation without requiring complete retraining. More broadly, this work investigates how anomaly-detection models can remain useful across different qubit domains.