Enhancing Variational Quantum Algorithms: Effective Quantum Ansatz Design Using GFlowNets

Jun Dai, Michael Rizvi-Martel, Guillaume Rabusseau

Keywords: quantum computing, GFlowNets, Classical-Quantum Hybrid Algorithms, Variational Quantum Algorithms (VQAs)

TL;DR: GFlowNets automate efficient quantum circuits discovery

Abstract: Quantum computing promises significant computational advantages over classical computing. However, current devices are constrained by a limited qubit count and noise. By combining classical optimization methods with parameterized quantum circuits, Variational Quantum Algorithms (VQAs) offer a potential solution for noisy intermediate-scale quantum systems (NISQ). This makes VQAs particularly promising strategies for achieving near-term quantum advantages; such approaches are now widely explored for nearly all quantum computing applications. However, designing effective parameterized circuits, also known as ansatz, remains challenging. In this work, we introduce the use of GFlowNets as an efficient method to automate the development of efficient ansatz for various quantum computing problems. Our approach leverages GFlowNets to efficiently explore the combinatorial space of parameterized quantum circuits. Our extensice experiments demonstrate that GFlowNets can discover ansatz with an order of magnitude fewer parameters, gate counts, and depths compared to current approaches for the molecular electronic ground state energy problem. We also apply our approach to the unweighted Max-Cut problem, where we observe similar improvements in circuit efficiency. These results highlight the potential of GFlowNets to significantly reduce the resource requirements of VQAs while maintaining or improving solution quality.

Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 11548