The increasing demand for computational power drives continuous development of large-scale quantum processors, but their fabrication remains expensive and limits accessibility. Wei-You Liao from Henan University and colleagues address this challenge by introducing predictive surrogates, which are classical learning models that efficiently emulate the behaviour of quantum processors. This innovative approach substantially reduces the need for direct access to these rare and costly machines, opening new avenues for research in areas such as materials science and fundamental physics. The team demonstrates these surrogates by successfully emulating a processor with 20 superconducting qubits, achieving significant reductions in measurement overhead and, crucially, surpassing the performance of traditional methods for complex calculations. These findings establish predictive surrogates as a practical means of extending the impact of advanced quantum computing beyond the limitations of current hardware availability.
The continuing development of quantum processors is driving breakthroughs in scientific discovery. Despite this progress, the formidable cost of fabricating large-scale quantum processors means they will remain rare for the foreseeable future, limiting their widespread application. To address this bottleneck, researchers introduce the concept of predictive surrogates, classical learning models designed to emulate the mean-value behaviour of a given quantum processor with provably computational efficiency. This work demonstrates their potential and outlines a path towards more efficient quantum computing workflows.
Paired Measurements with Four Decimal Precision
The data consists of paired numerical values, likely representing measurements or results from an experiment or simulation. Each row represents a distinct unit or sample, with two columns of associated data. The values range from approximately 0. 04 to 0. 85, presented to four decimal places, indicating a high degree of precision. With 79 rows of data, several analyses are possible, including calculating descriptive statistics, performing correlation analysis, and visualising relationships with scatter plots. Regression analysis could also model the relationship and predict values, while statistical tests could compare means if the data represents different groups.
Predictive Surrogates Emulate Quantum Computation Outcomes
Researchers have developed a new approach to broaden the impact of advanced quantum processors by creating “predictive surrogates”, classical machine learning models that accurately mimic the behaviour of quantum hardware. These surrogates address a significant bottleneck in the field, as access to increasingly powerful quantum computers remains limited due to their high fabrication costs and complex maintenance requirements. The team’s innovation allows researchers to perform complex calculations and pre-training tasks using classical computers, significantly reducing the need for direct access to quantum hardware. The predictive surrogates function by learning to estimate the average outcomes of quantum computations, effectively emulating the results a quantum processor would produce.
This is achieved through a training process where the surrogate is exposed to data generated by an actual quantum processor, allowing it to build an accurate model of the hardware’s behaviour. Importantly, these surrogates are designed with theoretical guarantees of efficiency, addressing a key limitation of previous machine learning approaches. Experimental results demonstrate that these surrogates can surpass the performance of conventional methods. The team successfully utilised them to pre-train variational quantum algorithms and identify complex quantum phases. This was achieved by reducing measurement overhead, enabling more efficient exploration of quantum systems and accelerating the pace of discovery, particularly in areas like quantum chemistry and materials science where complex simulations are essential. By providing a cost-effective and efficient alternative to direct quantum computation, these predictive surrogates promise to democratize access to advanced quantum capabilities and accelerate progress in the field.
Surrogate Models Expand Quantum Processor Access
This research introduces predictive surrogates, classical machine learning models designed to mimic the behaviour of quantum processors, offering a pathway to broaden the impact of these advanced, yet limited, resources. The team demonstrates that these surrogates can accurately emulate processors with up to twenty superconducting qubits, significantly reducing the need for direct access to the quantum hardware. This capability proves particularly valuable in complex tasks such as pre-training variational eigensolvers and identifying specific quantum phases of matter, including Floquet symmetry-protected topological phases. The results show that employing these surrogates not only reduces measurement overhead but also surpasses the performance of traditional, resource-intensive approaches to quantum computation.
By effectively predicting processor behaviour, the surrogates allow for more efficient optimisation and analysis, addressing a key bottleneck in the field. The authors acknowledge that the accuracy of the surrogate relies on the quality of the initial data obtained from the quantum processor, and that further research is needed to improve its generalizability across different processor architectures. Future work will focus on refining these models and exploring their application to a broader range of quantum algorithms and problems, ultimately aiming to make quantum computing more accessible and impactful.