Vqezy Dataset enables initialization of variable quantum eigenvalue parameters across 12,110 instances and 7 tasks

Variational quantum-specific solvers represent promising approaches to solving complex problems in short-term quantum computers, but their effectiveness depends on carefully selected initiation parameters. Chi Chang, Menk Singzen and Qian Lu from the University of Central Florida, together with Hui Min Leong and Fang Chen from Indiana University, addressed important bottlenecks in this field by creating a comprehensive open source data set called Vqezy. This new resource overcomes the limitations of existing datasets, which are often limited to specific problem types by providing over 12,000 instances across multiple domains and full optimization paths. Vqezy can develop and benchmark machine learning-based methods for researchers to initialize these algorithms, ultimately speeding up progress towards practical quantum computations, enabling more reliable results from noisy quantum hardware.

It achieves cutting-edge performance and the lack of comprehensive datasets limits its progress. Existing resources are typically restricted to a single domain and do not have full coverage of Hamiltonians, Unsatisfied Circuits, and optimization trajectories. To overcome these limitations, scientists developed VQezy, the first large dataset of VQE parameter initialization. The dataset is available on https://github. com/chizhang24/vqezy, and is continuously refined and expanded to promote advances in machine learning-based initialization methods.

Vqezy dataset for parameter initialization studies

Scientists have established Vqezy, a comprehensive large dataset designed to promote advances in variational quantum eigenvalue (VQE) parameter initialization. To build Vqezy, researchers systematically generated Hamiltonians, selected the appropriate Ansatz circuits, and performed VQE optimizations for each instance. The resulting dataset includes rich attributes for each instance, including Hamiltonian issues, detailed circuit specifications, and optimized VQE parameter vectors. For quantum many-body physics tasks, the team generated instances of the 1D Heisenberg XYZ model, the 1D Fermi Habbard model, and the 2D horizontal field ISING model, generating various parameters such as coupling constants and magnetic fields. Quantum chemistry task focusing on hydrogen molecules (H2) using a variety of basis sets and molecular geometry. This comprehensive approach allows researchers to explore diverse VQE configurations and develop more effective initialization strategies.

Vqezy dataset allows for large-scale parameter initialization

The research team established Vqezy. This is a comprehensive large dataset designed to facilitate advances in initialization of variational quantum eigenvalue (VQE) parameters. Vqezy consists of 12,110 instances spanning three major application areas: quantum many-body physics, quantum chemistry, and random VQE, providing a complete optimization trajectory for each instance. Within the domain of quantum many-body physics, the team generated data for the 1-dimensional Heisenberg XYZ model, the 1-dimensional Fermi Habbard model, and the 2-dimensional lateral field of view ISING model. For quantum chemistry applications, the dataset includes H2, HEH+, and NH3 configurations that are generated at different bond lengths.

The random VQE domain contributed four kit Hamiltonians with randomly generated coefficients. The team adopted T-Sisted Stochastic Neighbord Embedding (T-SNE) and Multidimensional Scaling (MDS) to visualize optimized VQE parameters, revealing different patterns within each application area. Openly available to the research community, this dataset is designed to be continuously refined and expanded, establishing the foundation for future VQE research.

Vqezy Dataset accelerates algorithm development

This paper presents Vqezy, a public dataset designed to accelerate research in the field of variational quantum eigenvalues ​​(VQE) algorithms. The authors identified the lack of large and diverse datasets that hinder VQE research, particularly in areas such as algorithm optimization and architectural design. Vqezy consists of 12,110 instances spanning three major VQE application domains, quantum chemistry, materials science, and physics. Researchers can use VQEZY to improve the efficiency and accuracy of VQE algorithms, develop algorithms that can be generalized across different quantum systems, and evaluate the performance of various VQE algorithms. The authors plan to expand the dataset to include data from larger molecules and more complex systems, as well as real quantum hardware.