QHyper: an integration library for hybrid quantum-classical optimization
Tomasz Lamża, Justyna Zawalska, Kacper Jurek, Mariusz Sterzel, Katarzyna Rycerz
TL;DR
The paper addresses the fragmentation of quantum optimization tooling by proposing QHyper, a unified Python library for hybrid quantum-classical combinatorial optimization. It provides a Polynomial-based problem representation, a Converter to convert problems into $QUBO$, $QuadraticConstrainedDiscreteOptimization$, and $QuadraticUnconstrainedDiscreteOptimization$ formulations, and interfaces to solvers such as $QAOA$ variants and D-Wave systems. It also offers local and global (hyper)optimizers and a YAML-driven configuration format for reproducible experimentation, including penalties $\alpha_j$ used in the objective. The contributions include a flexible, extensible architecture that supports adding custom problems, solvers, and optimizers, plus practical demonstrations and notebooks to facilitate benchmarking, rapid prototyping, and reproducible research in quantum optimization. The work enables researchers to systematically explore solver performance, hyperparameter effects, and potential quantum advantages, with significant implications for speeding up evaluation and adoption of hybrid quantum-classical approaches.
Abstract
We propose the QHyper library, which is aimed at researchers working on computational experiments with a variety of quantum combinatorial optimization solvers. The library offers a simple and extensible interface for formulating combinatorial optimization problems, selecting and running solvers, and optimizing hyperparameters. The supported solver set includes variational gate-based algorithms, quantum annealers, and classical solutions. The solvers can be combined with provided local and global (hyper)optimizers. The main features of the library are its extensibility on different levels of use as well as a straightforward and flexible experiment configuration format presented in the paper.