More symmetry in quantum models requires more measurement resources

Quantum kernels provide an effective procedure for learning quantum phase transitions on quantum processing devices, but scalability issues of learning strategies related to the symmetries of critical models remain to be resolved. Aaqib Ali and colleagues at the University of Bari have drawn a link between model symmetry and scaling of fidelity kernel resources. Their work quantified the measurement resources required to estimate a fidelity-based quantum kernel for many-body ground states and demonstrated that increasing the symmetry of the spin model systematically increases the number of shots required for accurate estimation. This discovery provides a practical, symmetry-aware limit for physics-based quantum machine learning and represents an important step toward efficient and reliable quantum algorithms for materials science and condensed matter physics.

Symmetry amplification exponentially increases quantum computational cost for kernel estimation

Switching from 2 From the symmetric Ising/XY model -Symmetry XX (and XXZ) models significantly increase the number of required quantum “shots”. Previously, accurately estimating fidelity-based quantum kernels required an undefined number of shots, limiting their practical application. This lack of clarity has hindered the development of scalable quantum machine learning algorithms for complex physical systems. The increased symmetry directly amplifies these shot requirements, reaching a threshold where accurate kernel estimation was previously difficult due to exponential resource demands. Fidelity-based quantum kernels essentially measure the overlap between two quantum states, and their accurate estimation is critical for tasks such as classification and regression within quantum machine learning frameworks. The computational cost associated with this estimation is directly related to the number of measurements or “shots” performed on the quantum processor.

Dr. Alan Diosi from the University of Szeged and Dr. Pan Zhang from the University of Science and Technology of China used the SWAP test estimator to quantify this effect and reveal a direct correlation between spin model symmetry and the computational resources required for quantum machine learning. The SWAP test is a quantum algorithm used to estimate the dot product between two quantum states and is directly related to the fidelity kernel. This estimator allows efficient computation of the kernel matrix elements, but its resource requirements are sensitive to the symmetry of the system. This correlation enables practical, symmetry-aware boundaries for physics-based quantum machine learning. Closed-form fidelity calculations were adopted for the free fermion XY and XX models, and their analytical solvability was utilized to obtain accurate results. These calculations served as a benchmark for more complex XXZ chains, where exact diagonalization was used to benchmark shot noise effects. Although exact diagonalization is computationally expensive, it provides a reliable method for determining the ground state of the XXZ chain and allows accurate evaluation of kernel estimation accuracy under different levels of noise.

The researchers found that the required shot scaling is not simply linear with increasing symmetry, but rather exhibits an exponential relationship. This means that even a small increase in symmetry can significantly increase the computational resources required to accurately estimate the fidelity kernel. For example, moving from a system with relatively low symmetry, such as the Ising model, to a system with higher symmetry, such as the XXZ model, may require an exponential increase in the number of quantum shots. This poses a significant challenge when implementing quantum machine learning algorithms in near-term quantum devices, which are limited in the number of qubits and coherence time available to perform computations.

The impact of symmetry on computational cost limits the accuracy of identifying changes in material properties

Despite the clarity of resource scaling, slow convergence rates remain a practical hurdle in identifying quantum phase transitions. Accurately identifying critical points, where material properties change dramatically, requires unrealistically large system sizes, even with advanced analysis techniques such as finite size scaling. Finite size scaling is a method used to estimate system behavior from finite size simulations to thermodynamic limits, but relies on accurate data at multiple system sizes. The increased computational cost associated with highly symmetric models limits the achievable system size and impedes the accuracy of finite-size scaling analyses. Therefore, algorithms need to be improved to better exploit symmetries and reduce computational complexity by favoring simpler models when necessary. Exploring alternative kernel estimators or developing techniques that take advantage of system symmetries may provide potential means to reduce computational load.

Identifying subtle changes in material properties with high accuracy remains a major challenge even with efficient estimators. The ability to accurately detect these changes is critical to designing and discovering new materials with customized properties. The direct relationship between the symmetries of quantum systems and the computational cost of machine learning is an important step forward for the field. Symmetries within the spin model systematically increase the number of measurements called “shots” required for accurate kernel estimation. Kernels serve as mathematical comparisons of quantum states. The researchers used exact diagonalization to benchmark the effects of shot noise, and their analysis revealed how symmetry affects the resource demands of fidelity-based quantum kernels used to identify quantum phase transitions. Its implications extend to various fields of condensed matter physics, including the study of magnetism, superconductivity, and topological phases of matter. Understanding the relationship between symmetry and computational cost is essential for developing efficient quantum algorithms to simulate and analyze these complex systems.

This study highlights the importance of considering the symmetries of the underlying physical model when designing quantum machine learning algorithms. Incorporating symmetry-aware boundaries into the learning process has the potential to reduce the required computational resources and improve the accuracy of the results. This work provides a valuable foundation for future research in quantum machine learning and opens new possibilities for exploring the properties of complex materials using quantum computers. Quantum computing materials shows that a careful balance between model complexity and computational feasibility is critical to achieving practical quantum benefits in materials science and condensed matter physics.

This study establishes a clear relationship between the symmetry of quantum spin models and the number of measurements required for accurate machine learning. This is important because symmetry improvements require significantly more computational “shots”, with up to a significant increase when moving from simple models to more complex models like XXZ chains, impeding the scalability of quantum algorithms. Researchers used techniques such as exact diagonalization and fidelity-based kernels to quantify this relationship and reveal how symmetry affects resource demand. This understanding could lead to the development of more efficient quantum machine learning algorithms tailored to the symmetries of specific materials, ultimately accelerating materials discovery and the study of complex quantum systems.