Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

arXiv:2205.06217 · PRX Quantum

Exploiting Symmetry in Variational Quantum Machine Learning

Johannes Jakob Meyer, Marian Mularski, Elies Gil-Fuster, Antonio Anna Mele, Francesco Arzani, Alissa Wilms, Jens Eisert

Published Mar 17, 2023• 172 citations
arXivHTMLPDF

Abstract

A blueprint for exploiting symmetries in the construction of variational quantum learning models that can result in improved generalization performance is developed and demonstrated on practical problems.
DOI
10.1103/prxquantum.4.010328

Related papers

Variational Quantum Algorithms and Quantum Machine Learning

Kosuke Mitarai

電子情報通信学会誌•Nov 1, 2021•0 citations

Resource-efficient equivariant quantum convolutional neural networks

Koki Chinzei, Quoc Hoan Tran, Yasuhiro Endo +1 more

Abstract Equivariant quantum neural networks (QNNs) are promising variational models that exploit symmetries to improve machine learning capabilities. Despite theoretical developments in equivariant QNNs, their implementation on near-term quantum devices remains challenging due to limited computational resources. This study proposes a resource-efficient model of equivariant quantum convolutional networks (QCNNs) called equivariant split-parallelizing QCNN (sp-QCNN). Using a group-theoretical approach, we encode general symmetries into our model beyond the translational symmetry addressed by previous sp-QCNNs. We achieve this by splitting the circuit at the pooling layer while preserving symmetry. This splitting structure effectively parallelizes QCNNs to improve measurement efficiency in estimating the expectation value of an observable and its gradient by order of the number of qubits. Our model also exhibits high trainability and generalization performance, including the absence of barren plateaus. Numerical experiments demonstrate that the equivariant sp-QCNN can be trained and generalized with fewer measurement resources than a conventional equivariant QCNN in a noisy quantum data classification task. Our results contribute to the advancement of practical quantum machine learning algorithms.

arXiv:2410.01252•Quantum Machine Intelligence•Apr 27, 2026•1 citation

Kvanttikoneoppimismallien toteutus ongelmien ratkomisesta

Ilmo Salmenperä

Quantum Machine Learning (QML), the intersection of quantum computing and machine learning, has been thought to be one of the more promising ways to achieve quantum advantage in the immediate future. Applying these well-defined models in practical settings comes with various restrictions that are dependent on either the underlying hardware or the mathematical formalization of the model itself. In this thesis, we study these two classes of issues, find solutions to them in different contexts, and then generalize the common elements these issues and solutions have. In the case of hardware-related issues, the key observation is that the hardware features, like qubit counts, gate topology, or error rates, tend to limit the size of the problem we want to solve. These issues can usually be solved by optimizing the model's structure to make it more compatible with the hardware on which it is run. This holds with gate-based approaches, like how we used quantum compilation routines to optimize an existing quantum linear regression routine to be runnable in existing quantum hardware, but also with quantum annealing hardware, like how we leveraged classical software techniques to improve the efficiency of embedding a restricted Boltzmann machine model to a quantum annealing device. Issues arising from the model construction process tend to be less uniform and are highly dependent on the model itself. The model formalization can give rise to unintended phenomena, which can be alleviated either by accounting for them during use or by adjusting the model itself to negate the issue. We found examples of these issues emerging in existing QML models, such as the phenomenon of self-erasing gates in the established way of constructing quantum embedding kernels, and how the permutation of features can have a considerable impact on the performance of variational QML models. We also developed solutions to these issues in the form of improvements on the models themselves or as best practices on how to circumvent them. While the research topic of this thesis is incredibly vast, our research is a step towards more practical QML models. It encompasses many distinct QML models and hardware technologies on which these models can be run. By developing insight into how QML algorithms operate in practical settings, we also develop readiness for a possible future in which quantum computers are a viable option for machine learning tasks.

Työväentutkimus Vuosikirja•Dec 5, 2025•0 citations

Compositional Quantum Heuristics for Max-Clique Detection

Tiffany Duneau, Colin Krawchuk, Anna Pearson

Quantum machine learning holds the promise of combining the success of classical machine learning methods with the power of quantum computing, however one of the largest obstacles facing the field is the problem of barren plateaus. Parameterised quantum circuits offer a flexible framework for developing quantum machine learning models, but their practicality is constrained by a trade-off between trainability and classical simulability. In general, circuits that are sufficiently expressive to model complex behaviour often exhibit barren plateaus, where gradients vanish and optimisation fails. In this work we investigate a compositional approach to mitigate this trade-off by assembling larger quantum models from smaller subcomponents. To ensure trainability of these subcomponents, we describe a framework for constructing group-invariant loss functions, which introduce symmetry-induced inductive bias and lead to improved gradient behaviour and generalisation. In particular, we use this framework to design permutation-equivariant quantum graph neural networks for identifying maximal cliques in graphs. The models we construct exhibit superior training gradients through symmetry-induced bias, and our experiments demonstrate that the trained models generalise to larger, more complex problem instances. Finally, inspired by Quantum-Informed Recursive Optimisation Algorithms (arXiv:2308.13607), we implement a recursive hybrid quantum-classical heuristic using the learned quantum models to guide a classical search procedure, demonstrating improved inference accuracy and scalability. Together, these results suggest that compositional circuits could be a viable pathway towards scalable quantum learning models that remain challenging to reproduce classically.

quant-pharXiv:2605.07611•Quantum Science and Technology•May 11, 2026•0 citations

An Example of Use of Variational Methods in Quantum Machine Learning

Marco Simonetti, Damiano Perri, Osvaldo Gervasi

arXiv:2208.04316•Lecture notes in computer science•Jan 1, 2022•5 citations

Quantum machine learning for particle physics using a variational quantum classifier

Blance, Andrew, Spannowsky, Michael

•Feb 24, 2021•0 citations
Abstract Equivariant quantum neural networks (QNNs) are promising variational models that exploit symmetries to improve machine learning capabilities. Despite theoretical developments in equivariant QNNs, their implementation on near-term quantum devices remains challenging due to limited computational resources. This study proposes a resource-efficient model of equivariant quantum convolutional networks (QCNNs) called equivariant split-parallelizing QCNN (sp-QCNN). Using a group-theoretical approach, we encode general symmetries into our model beyond the translational symmetry addressed by previous sp-QCNNs. We achieve this by splitting the circuit at the pooling layer while preserving symmetry. This splitting structure effectively parallelizes QCNNs to improve measurement efficiency in estimating the expectation value of an observable and its gradient by order of the number of qubits. Our model also exhibits high trainability and generalization performance, including the absence of barren plateaus. Numerical experiments demonstrate that the equivariant sp-QCNN can be trained and generalized with fewer measurement resources than a conventional equivariant QCNN in a noisy quantum data classification task. Our results contribute to the advancement of practical quantum machine learning algorithms.
Quantum Machine Learning (QML), the intersection of quantum computing and machine learning, has been thought to be one of the more promising ways to achieve quantum advantage in the immediate future. Applying these well-defined models in practical settings comes with various restrictions that are dependent on either the underlying hardware or the mathematical formalization of the model itself. In this thesis, we study these two classes of issues, find solutions to them in different contexts, and then generalize the common elements these issues and solutions have. In the case of hardware-related issues, the key observation is that the hardware features, like qubit counts, gate topology, or error rates, tend to limit the size of the problem we want to solve. These issues can usually be solved by optimizing the model's structure to make it more compatible with the hardware on which it is run. This holds with gate-based approaches, like how we used quantum compilation routines to optimize an existing quantum linear regression routine to be runnable in existing quantum hardware, but also with quantum annealing hardware, like how we leveraged classical software techniques to improve the efficiency of embedding a restricted Boltzmann machine model to a quantum annealing device. Issues arising from the model construction process tend to be less uniform and are highly dependent on the model itself. The model formalization can give rise to unintended phenomena, which can be alleviated either by accounting for them during use or by adjusting the model itself to negate the issue. We found examples of these issues emerging in existing QML models, such as the phenomenon of self-erasing gates in the established way of constructing quantum embedding kernels, and how the permutation of features can have a considerable impact on the performance of variational QML models. We also developed solutions to these issues in the form of improvements on the models themselves or as best practices on how to circumvent them. While the research topic of this thesis is incredibly vast, our research is a step towards more practical QML models. It encompasses many distinct QML models and hardware technologies on which these models can be run. By developing insight into how QML algorithms operate in practical settings, we also develop readiness for a possible future in which quantum computers are a viable option for machine learning tasks.
Quantum machine learning holds the promise of combining the success of classical machine learning methods with the power of quantum computing, however one of the largest obstacles facing the field is the problem of barren plateaus. Parameterised quantum circuits offer a flexible framework for developing quantum machine learning models, but their practicality is constrained by a trade-off between trainability and classical simulability. In general, circuits that are sufficiently expressive to model complex behaviour often exhibit barren plateaus, where gradients vanish and optimisation fails. In this work we investigate a compositional approach to mitigate this trade-off by assembling larger quantum models from smaller subcomponents. To ensure trainability of these subcomponents, we describe a framework for constructing group-invariant loss functions, which introduce symmetry-induced inductive bias and lead to improved gradient behaviour and generalisation. In particular, we use this framework to design permutation-equivariant quantum graph neural networks for identifying maximal cliques in graphs. The models we construct exhibit superior training gradients through symmetry-induced bias, and our experiments demonstrate that the trained models generalise to larger, more complex problem instances. Finally, inspired by Quantum-Informed Recursive Optimisation Algorithms (arXiv:2308.13607), we implement a recursive hybrid quantum-classical heuristic using the learned quantum models to guide a classical search procedure, demonstrating improved inference accuracy and scalability. Together, these results suggest that compositional circuits could be a viable pathway towards scalable quantum learning models that remain challenging to reproduce classically.