Skip to main content
Publication

Generative model benchmarks for superconducting qubits

by Kathleen E Hamilton, Eugen F Dumitrescu, Raphael C Pooser
Publication Type
Journal
Journal Name
Physical Review A
Publication Date
Page Number
062323
Volume
99
Issue
6

In this work we experimentally demonstrate how cloud-based deployment of non-adversarial gen- erative model training can be used as a benchmark for small (< 5 qubits) quantum devices. Using the 2 × 2 Bars and Stripes dataset, we demonstrate that incorporating hardware connectivity con- siderations into circuit ansatz design improves generative modeling with superconducting qubits. Performance is quantified using three commonly used metrics in data analytics: the Kullbeck-Leiber divergence, and two adaptations of the F1 score.