Noisy circuit training of generative models with superconducting qubits

Presenting Author: Kathleen Hamilton, Oak Ridge National Laboratory
Contributing Author(s): Eugene Dumitrescu, Raphael Pooser

Many NISQ devices with < 20 qubits are becoming available for public use, but the lack of detailed noise models makes it difficult to use simulation to predict performance on hardware. We have used a recently introduced class of generative models [1] to quantify the performance of noisy superconducting qubits. Many sources of error and noise can be identified (e.g. decoherence, gate fidelities and measurement errors), and while the use of noise-robust stochastic optimizers can train circuits to a reasonable degree of accuracy, the cohesive incorporation of error mitigation into circuit training remains an open question. Our work focuses on how the performance of generative models on noisy qubits can be improved without error mitigation: by minimizing the number of noisy gates in a circuit and using sampling rates to improve the dynamics of gradient-based circuit training. [1] Liu, Jin-Guo, and Lei Wang. "Differentiable learning of quantum circuit Born machine." arXiv:1804.04168 (2018). This work was supported as part of the ASCR Testbed Pathfinder Program at Oak Ridge National Laboratory under FWP #ERKJ332

(Session 7 : Monday from 11:30am - 12:00pm)


SQuInT Chief Organizer
Akimasa Miyake, Associate Professor

SQuInT Local Organizers
Rafael Alexander, Postdoctoral Fellow
Chris Jackson, Postdoctoral Fellow

SQuInT Administrator
Gloria Cordova
505 277-1850

SQuInT Assistant
Wendy Jay

SQuInT Founder
Ivan Deutsch, Regents' Professor, CQuIC Director

Tweet About SQuInT 2019!