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:2405.00252

Q-Newton: Hybrid Quantum-Classical Scheduling for Accelerating Neural Network Training with Newton's Gradient Descent

Pingzhi Li, Junyu Liu, Hanrui Wang, Tianlong Chen

Published Apr 30, 2024• 2 citations
arXivHTMLPDF

Abstract

Optimization techniques in deep learning are predominantly led by first-order gradient methodologies, such as SGD. However, neural network training can greatly benefit from the rapid convergence characteristics of second-order optimization. Newton's GD stands out in this category, by rescaling the gradient using the inverse Hessian. Nevertheless, one of its major bottlenecks is matrix inversion, which is notably time-consuming in $O(N^3)$ time with weak scalability. Matrix inversion can be translated into solving a series of linear equations. Given that quantum linear solver algorithms (QLSAs), leveraging the principles of quantum superposition and entanglement, can operate within a $\text{polylog}(N)$ time frame, they present a promising approach with exponential acceleration. Specifically, one of the most recent QLSAs demonstrates a complexity scaling of $O(d\cdotκ\log(N\cdotκ/ε))$, depending on: {size~$N$, condition number~$κ$, error tolerance~$ε$, quantum oracle sparsity~$d$} of the matrix. However, this also implies that their potential exponential advantage may be hindered by certain properties (i.e. $κ$ and $d$). We propose Q-Newton, a hybrid quantum-classical scheduler for accelerating neural network training with Newton's GD. Q-Newton utilizes a streamlined scheduling module that coordinates between quantum and classical linear solvers, by estimating & reducing $κ$ and constructing $d$ for the quantum solver. Our evaluation showcases the potential for Q-Newton to significantly reduce the total training time compared to commonly used optimizers like SGD. We hypothesize a future scenario where the gate time of quantum machines is reduced, possibly realized by attoseconds physics. Our evaluation establishes an ambitious and promising target for the evolution of quantum computing.

DOI
10.48550/arxiv.2405.00252

Related papers

A quantum algorithm for training wide and deep classical neural networks

Alexander Zlokapa, Hartmut Neven, Seth Lloyd

Given the success of deep learning in classical machine learning, quantum algorithms for traditional neural network architectures may provide one of the most promising settings for quantum machine learning. Considering a fully-connected feedforward neural network, we show that conditions amenable to classical trainability via gradient descent coincide with those necessary for efficiently solving quantum linear systems. We propose a quantum algorithm to approximately train a wide and deep neural network up to $O(1/n)$ error for a training set of size $n$ by performing sparse matrix inversion in $O(\log n)$ time. To achieve an end-to-end exponential speedup over gradient descent, the data distribution must permit efficient state preparation and readout. We numerically demonstrate that the MNIST image dataset satisfies such conditions; moreover, the quantum algorithm matches the accuracy of the fully-connected network. Beyond the proven architecture, we provide empirical evidence for $O(\log n)$ training of a convolutional neural network with pooling.

arXiv:2107.09200•Jul 19, 2021•23 citations

Hybrid Quantum Classical Acceleration of Machine Learning Algorithms: A Performance Study

Mahesh V, Ramkumar Krishnamoorthy, Diwakara Vasuman M S +2 more

The recent rise in the number of data-intensive applications has revealed some inherent limitations of traditional machine learning, such as high training costs, slow convergence rates, and the uncapable of scalability of traditional machine learning in high-dimensional spaces. Bottlenecks are caused by the iterative optimization procedure, kernel evaluations, and similarity computations that are not efficiently supported by traditional hardware. Although recent advances in quantum computing, such as the Variational Quantum Algorithm, quantum kernels, and parameterized quantum circuits, hold promise for acceleration, the application of quantum speedup is presently hindered by noisy intermediate-scale quantum computing. This research provides a hybrid quantum-classical acceleration method that integrates quantum methods with selected computationally intensive components of classical machine learning systems rather than attempting to entirely replace them. Carried out a comprehensive analysis of the performance of well-known algorithms like Support Vector Machines, k-means clustering, and gradient-based neural networks in classical and quantum-assisted hybrid scenarios. The quantum resources are greatly beneficial for particular sub-tasks such as optimization, approximation of similarity, and exploration of the feature space, with the remaining learning process performed classically. Our analysis reveals that hybrid methods can offer a tangible acceleration and improved convergence behaviour in particular scenarios without sacrificing accuracy. Moreover, the analysis investigates the trade-offs associated with the utilization of quantum resources, robustness to noise, and orchestration costs.

EPJ Web of Conferences•Jan 1, 2026•0 citations

Hybrid Quantum Classical Acceleration of Machine Learning Algorithms: A Performance Study

Mahesh V, Ramkumar Krishnamoorthy, Diwakara Vasuman M S +2 more

The recent rise in the number of data-intensive applications has revealed some inherent limitations of traditional machine learning, such as high training costs, slow convergence rates, and the uncapable of scalability of traditional machine learning in high-dimensional spaces. Bottlenecks are caused by the iterative optimization procedure, kernel evaluations, and similarity computations that are not efficiently supported by traditional hardware. Although recent advances in quantum computing, such as the Variational Quantum Algorithm, quantum kernels, and parameterized quantum circuits, hold promise for acceleration, the application of quantum speedup is presently hindered by noisy intermediate-scale quantum computing. This research provides a hybrid quantum-classical acceleration method that integrates quantum methods with selected computationally intensive components of classical machine learning systems rather than attempting to entirely replace them. Carried out a comprehensive analysis of the performance of well-known algorithms like Support Vector Machines, k-means clustering, and gradient-based neural networks in classical and quantum-assisted hybrid scenarios. The quantum resources are greatly beneficial for particular sub-tasks such as optimization, approximation of similarity, and exploration of the feature space, with the remaining learning process performed classically. Our analysis reveals that hybrid methods can offer a tangible acceleration and improved convergence behaviour in particular scenarios without sacrificing accuracy. Moreover, the analysis investigates the trade-offs associated with the utilization of quantum resources, robustness to noise, and orchestration costs.

•Mar 23, 2026•0 citations

Quantum-Inspired Optimization of Large-Scale Machine Learning Models

Jincheng Zhang

The training of large-scale machine learning models presents significant computational challenges, primarily due to the inherent complexity of optimization algorithms like stochastic gradient descent (SGD) and its variants. These algorithms often struggle with convergence speed, getting trapped in local minima, and the ability to effectively explore the vast solution space. This paper investigates the potential of quantum-inspired optimization techniques to address these limitations. Specifically, we explore the application of quantum annealing and variational quantum eigensolver (VQE) approaches to accelerate the training and optimization process of large-scale machine learning models. Our core claim is that by mapping the optimization problem to a quantum system, we can leverage quantum computation to overcome the computational bottlenecks associated with classical optimization methods. We propose a framework where the objective function of the machine learning model is translated into a Hamiltonian for a quantum annealer or the cost function for a VQE, enabling a quantum-inspired solution. The results, while theoretical in this initial exploration, demonstrate a potential pathway for significantly reducing training times and improving the overall performance of large-scale machine learning models. Further research is needed to develop efficient hybrid quantum-classical algorithms and explore the practical limitations of this approach.

•Sep 6, 2026•0 citations

Quantum Gradient Descent in Neural Network Training Optimization

Zacharias Rallis, Kyriakos Vougiouklis

Optimisation of training still faces challenges due to an ill-conditioned loss surface and noisy gradient signals; at the same time, limited computational resources also restrict training options. Quantum Gradient Descent as a Hybrid Optimisation Strategy for Neural Network Training. The proposed noise-resilient quantum gradient descent framework maps selected gradient components to parameterised quantum circuits, estimates directional derivatives via parameter-shift measurements, and returns stable updates to the classical optimizer. It is not directly substituted for backpropagation; instead, it is a type of controllable gradient module that can be added to sensitive layers or low-dimensional projection heads. Add adaptive shot scheduling, damped momentum and layer-wise trust constraints to reduce estimator variance and prevent unstable quantum updates. Representative classification tasks are selected as the subjects of experiments for convergence speed, validation accuracy, gradient fidelity, measurement cost and noise robustness. Based on the above analysis, the new optimizer outperforms standard SGD in terms of final accuracy by 1.6% - 3.2% when using the same number of epochs; at the same time, it reduces the median gradient variance by 28.4% and maintains stable convergence even with an increase in simulated depolarizing noise from 0.001 to 0.015. Based on the above, it can be seen that Quantum Gradient Descent is not an all-purpose optimizer but rather a target-specific training accelerator for gradient-sensitive modules that can trade off sampling cost for more informative descent directions in measurement-aware update design.

Journal of Applied Automation Technologies•Aug 24, 2023•0 citations

Optimization of Neural Networks Using Quantum Computing Techniques

Ayesha Javed, Muhammad Arqam Raza, Sana Khalid

Quantum computing offers new algorithmic primitives—superposition, entanglement, and interference—that can accelerate or qualitatively change parts of the machine-learning pipeline. This article surveys optimization of neural networks using quantum techniques, focusing on three avenues: (i) quantum-accelerated linear algebra and sampling to speed up gradient and Hessian-vector computations; (ii) hybrid variational algorithms where parameterized quantum circuits (PQCs) act as feature maps or layers within classical networks; and (iii) quantum-aware optimizers such as the quantum natural gradient (QNG) that respect the geometry of quantum states. We synthesize theoretical results with NISQ-era constraints (noise, barren plateaus, trainability), propose design patterns for end-to-end systems, and present an illustrative comparison of convergence between classical SGD and a hybrid QNG setup. The analysis indicates that, for structured data and appropriately shallow, problem-matched ansätze, hybrid methods can improve sample efficiency and generalization while remaining hardware-feasible.

International Journal of Advanced Sciences and Computing•Jun 30, 2022•0 citations
Optimization techniques in deep learning are predominantly led by first-order gradient methodologies, such as SGD. However, neural network training can greatly benefit from the rapid convergence characteristics of second-order optimization. Newton's GD stands out in this category, by rescaling the gradient using the inverse Hessian. Nevertheless, one of its major bottlenecks is matrix inversion, which is notably time-consuming in O(N3)O(N^3)O(N3) time with weak scalability. Matrix inversion can be translated into solving a series of linear equations. Given that quantum linear solver algorithms (QLSAs), leveraging the principles of quantum superposition and entanglement, can operate within a polylog(N)\text{polylog}(N)polylog(N) time frame, they present a promising approach with exponential acceleration. Specifically, one of the most recent QLSAs demonstrates a complexity scaling of O(d⋅κlog⁡(N⋅κ/ε))O(d\cdotκ\log(N\cdotκ/ε))O(d⋅κlog(N⋅κ/ε)), depending on: {size~NNN, condition number~κκκ, error tolerance~εεε, quantum oracle sparsity~ddd} of the matrix. However, this also implies that their potential exponential advantage may be hindered by certain properties (i.e. κκκ and ddd). We propose Q-Newton, a hybrid quantum-classical scheduler for accelerating neural network training with Newton's GD. Q-Newton utilizes a streamlined scheduling module that coordinates between quantum and classical linear solvers, by estimating & reducing κκκ and constructing ddd for the quantum solver. Our evaluation showcases the potential for Q-Newton to significantly reduce the total training time compared to commonly used optimizers like SGD. We hypothesize a future scenario where the gate time of quantum machines is reduced, possibly realized by attoseconds physics. Our evaluation establishes an ambitious and promising target for the evolution of quantum computing.
Given the success of deep learning in classical machine learning, quantum algorithms for traditional neural network architectures may provide one of the most promising settings for quantum machine learning. Considering a fully-connected feedforward neural network, we show that conditions amenable to classical trainability via gradient descent coincide with those necessary for efficiently solving quantum linear systems. We propose a quantum algorithm to approximately train a wide and deep neural network up to O(1/n)O(1/n)O(1/n) error for a training set of size nnn by performing sparse matrix inversion in O(log⁡n)O(\log n)O(logn) time. To achieve an end-to-end exponential speedup over gradient descent, the data distribution must permit efficient state preparation and readout. We numerically demonstrate that the MNIST image dataset satisfies such conditions; moreover, the quantum algorithm matches the accuracy of the fully-connected network. Beyond the proven architecture, we provide empirical evidence for O(log⁡n)O(\log n)O(logn) training of a convolutional neural network with pooling.
The recent rise in the number of data-intensive applications has revealed some inherent limitations of traditional machine learning, such as high training costs, slow convergence rates, and the uncapable of scalability of traditional machine learning in high-dimensional spaces. Bottlenecks are caused by the iterative optimization procedure, kernel evaluations, and similarity computations that are not efficiently supported by traditional hardware. Although recent advances in quantum computing, such as the Variational Quantum Algorithm, quantum kernels, and parameterized quantum circuits, hold promise for acceleration, the application of quantum speedup is presently hindered by noisy intermediate-scale quantum computing. This research provides a hybrid quantum-classical acceleration method that integrates quantum methods with selected computationally intensive components of classical machine learning systems rather than attempting to entirely replace them. Carried out a comprehensive analysis of the performance of well-known algorithms like Support Vector Machines, k-means clustering, and gradient-based neural networks in classical and quantum-assisted hybrid scenarios. The quantum resources are greatly beneficial for particular sub-tasks such as optimization, approximation of similarity, and exploration of the feature space, with the remaining learning process performed classically. Our analysis reveals that hybrid methods can offer a tangible acceleration and improved convergence behaviour in particular scenarios without sacrificing accuracy. Moreover, the analysis investigates the trade-offs associated with the utilization of quantum resources, robustness to noise, and orchestration costs.
The recent rise in the number of data-intensive applications has revealed some inherent limitations of traditional machine learning, such as high training costs, slow convergence rates, and the uncapable of scalability of traditional machine learning in high-dimensional spaces. Bottlenecks are caused by the iterative optimization procedure, kernel evaluations, and similarity computations that are not efficiently supported by traditional hardware. Although recent advances in quantum computing, such as the Variational Quantum Algorithm, quantum kernels, and parameterized quantum circuits, hold promise for acceleration, the application of quantum speedup is presently hindered by noisy intermediate-scale quantum computing. This research provides a hybrid quantum-classical acceleration method that integrates quantum methods with selected computationally intensive components of classical machine learning systems rather than attempting to entirely replace them. Carried out a comprehensive analysis of the performance of well-known algorithms like Support Vector Machines, k-means clustering, and gradient-based neural networks in classical and quantum-assisted hybrid scenarios. The quantum resources are greatly beneficial for particular sub-tasks such as optimization, approximation of similarity, and exploration of the feature space, with the remaining learning process performed classically. Our analysis reveals that hybrid methods can offer a tangible acceleration and improved convergence behaviour in particular scenarios without sacrificing accuracy. Moreover, the analysis investigates the trade-offs associated with the utilization of quantum resources, robustness to noise, and orchestration costs.
The training of large-scale machine learning models presents significant computational challenges, primarily due to the inherent complexity of optimization algorithms like stochastic gradient descent (SGD) and its variants. These algorithms often struggle with convergence speed, getting trapped in local minima, and the ability to effectively explore the vast solution space. This paper investigates the potential of quantum-inspired optimization techniques to address these limitations. Specifically, we explore the application of quantum annealing and variational quantum eigensolver (VQE) approaches to accelerate the training and optimization process of large-scale machine learning models. Our core claim is that by mapping the optimization problem to a quantum system, we can leverage quantum computation to overcome the computational bottlenecks associated with classical optimization methods. We propose a framework where the objective function of the machine learning model is translated into a Hamiltonian for a quantum annealer or the cost function for a VQE, enabling a quantum-inspired solution. The results, while theoretical in this initial exploration, demonstrate a potential pathway for significantly reducing training times and improving the overall performance of large-scale machine learning models. Further research is needed to develop efficient hybrid quantum-classical algorithms and explore the practical limitations of this approach.
Optimisation of training still faces challenges due to an ill-conditioned loss surface and noisy gradient signals; at the same time, limited computational resources also restrict training options. Quantum Gradient Descent as a Hybrid Optimisation Strategy for Neural Network Training. The proposed noise-resilient quantum gradient descent framework maps selected gradient components to parameterised quantum circuits, estimates directional derivatives via parameter-shift measurements, and returns stable updates to the classical optimizer. It is not directly substituted for backpropagation; instead, it is a type of controllable gradient module that can be added to sensitive layers or low-dimensional projection heads. Add adaptive shot scheduling, damped momentum and layer-wise trust constraints to reduce estimator variance and prevent unstable quantum updates. Representative classification tasks are selected as the subjects of experiments for convergence speed, validation accuracy, gradient fidelity, measurement cost and noise robustness. Based on the above analysis, the new optimizer outperforms standard SGD in terms of final accuracy by 1.6% - 3.2% when using the same number of epochs; at the same time, it reduces the median gradient variance by 28.4% and maintains stable convergence even with an increase in simulated depolarizing noise from 0.001 to 0.015. Based on the above, it can be seen that Quantum Gradient Descent is not an all-purpose optimizer but rather a target-specific training accelerator for gradient-sensitive modules that can trade off sampling cost for more informative descent directions in measurement-aware update design.
Quantum computing offers new algorithmic primitives—superposition, entanglement, and interference—that can accelerate or qualitatively change parts of the machine-learning pipeline. This article surveys optimization of neural networks using quantum techniques, focusing on three avenues: (i) quantum-accelerated linear algebra and sampling to speed up gradient and Hessian-vector computations; (ii) hybrid variational algorithms where parameterized quantum circuits (PQCs) act as feature maps or layers within classical networks; and (iii) quantum-aware optimizers such as the quantum natural gradient (QNG) that respect the geometry of quantum states. We synthesize theoretical results with NISQ-era constraints (noise, barren plateaus, trainability), propose design patterns for end-to-end systems, and present an illustrative comparison of convergence between classical SGD and a hybrid QNG setup. The analysis indicates that, for structured data and appropriately shallow, problem-matched ansätze, hybrid methods can improve sample efficiency and generalization while remaining hardware-feasible.