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:2602.00783

Analysis of Hessian Scaling for Local and Global Costs in Variational Quantum Algorithm

Yihan Huang, Yangshuai Wang

quant-phmath.NASubmitted Jan 31, 2026• Announced Feb 3, 2026• 0 citations• Revised Feb 10, 2026 (v2)
v1 · Jan 31, 2026v2 · Feb 10, 2026
arXivHTMLPDF

Abstract

Barren plateaus in variational quantum algorithms are typically described by gradient concentration at random initialization. In contrast, rigorous results for the Hessian, even at the level of entry-wise variance, remain limited. In this work, we analyze the scaling of Hessian-entry variances at initialization. Using exact second-order parameter-shift identities, we write $H_{jk}$ as a constant-size linear combination of shifted cost evaluations, which reduces ${\rm Var}_ρ(H_{jk})$ to a finite-dimensional covariance--quadratic form. For global objectives, under an exponential concentration condition on the cost at initialization, ${\rm Var}_ρ(H_{jk})$ decays exponentially with the number of qubits $n$. For local averaged objectives in bounded-depth circuits, ${\rm Var}_ρ(H_{jk})$ admits polynomial bounds controlled by the growth of the backward lightcone on the interaction graph. As a consequence, the number of measurement shots required to estimate $H_{jk}$ to fixed accuracy inherits the same exponential (global) or polynomial (local) scaling. Extensive numerical experiments over system size, circuit depth, and interaction graphs validate the predicted variance scaling. Overall, the paper quantifies when Hessian entries can be resolved at initialization under finite sampling, providing a mathematically grounded basis for second-order information in variational optimization.

AI Summary

The paper studies whether individual Hessian entries in variational quantum algorithms remain statistically resolvable at random initialization, contrasting global observables with averaged local costs. It considers independently parametrized gates with two-point-spectrum generators, product-uniform initialization, and bounded-depth or sublinearly growing-depth circuits. Exact second-order parameter-shift identities express each Hessian entry as at most four shifted cost evaluations. Its initialization variance is therefore a constant-dimensional covariance quadratic form. Shift invariance and Cauchy–Schwarz imply that an assumed cost-concentration bound $\operatorname{Var}(C)\le c_g e^{-\alpha n}$ transfers to every Hessian entry. For averaged term-wise $k$-local costs, a backward-lightcone dependency-graph argument gives $\operatorname{Var}(H_{jk})\le c_{\mathrm{loc}}V_G(k+2rL)/n$, where $V_G$ measures interaction-graph neighborhood growth. On $D$-dimensional lattices this becomes $O((k+2rL)^D/n)$ at fixed depth. Statevector simulations, primarily up to 16 qubits with 200 initializations per configuration, report global diagonal/off-diagonal variances approximately $e^{-0.57n}$ and $e^{-0.64n}$, versus local fits $n^{-1.96}$ and $n^{-2.05}$. Additional tests examine an Ising-model energy, Hessian spectra, and finite-shot behavior. Resolution is defined by shot noise falling below a fraction of initialization variance, with a sufficient shot budget proportional to its inverse. The paper interprets this as exponential versus polynomial resolution costs; however, the local theorem supplies only an upper variance bound, not the lower bound needed to guarantee polynomial resolvability. Stated limitations include instance-dependent exponents, conditional global concentration, no spectral-conditioning guarantees, and no device-noise analysis.

Likely largely AI-generatedIncrementalOverclaimsStrong assumptionsPossible error

Central claim may not hold: The paper incorrectly infers polynomial local Hessian-resolution shot requirements from a polynomial upper bound on Hessian-entry variance, without a corresponding signal lower bound. The abstract attributes exponential-versus-polynomial shot scaling to fixed-accuracy estimation, although its system-size-dependent criterion actually concerns resolution relative to initialization fluctuations. The local variance theorem is false as stated because repeated local terms on the same qubit satisfy its assumptions but can give constant Hessian variance instead of the claimed O(1/n) bound. The circuit assumptions leave entangler locality and within-layer depth unrestricted, so they do not establish the backward-lightcone bound used in the local proof.

Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.

Related papers

Optimized numerical gradient and Hessian estimation for variational quantum algorithms

Yong Siah Teo

Sampling noisy intermediate-scale quantum devices is a fundamental step that converts coherent quantum-circuit outputs to measurement data for running variational quantum algorithms that utilize gradient and Hessian methods in cost-function optimization tasks. This step, however, introduces estimation errors in the resulting gradient or Hessian computations. To minimize these errors, we discuss tunable numerical estimators, which are the finite difference (including their generalized versions) and scaled parameter-shift estimators [introduced in Phys. Rev. A 103, 012405 (2021)], and propose operational circuit-averaged methods to optimize them. We show that these optimized numerical estimators offer estimation errors that drop exponentially with the number of circuit qubits for a given sampling-copy number, revealing a direct compatibility with the barren-plateau phenomenon. In particular, there exists a critical sampling-copy number below which an optimized difference estimator gives a smaller average estimation error in contrast to the standard (analytical) parameter-shift estimator, which exactly computes gradient and Hessian components. Moreover, this critical number grows exponentially with the circuit-qubit number. Finally, by forsaking analyticity, we demonstrate that the scaled parameter-shift estimators beat the standard unscaled ones in estimation accuracy under any situation, with comparable performances to those of the difference estimators within significant copy-number ranges, and are the best ones if larger copy numbers are affordable.

arXiv:2206.12643•Physical Review A•Apr 17, 2023•3 citations

Unified Theoretical Characterization of Barren Plateaus in Variational Quantum Algorithms

Yalla Jnan Devi Satya Prasad

Barren plateaus-regions of exponentially small gradients-are a major obstacle for training variational quantum algorithms on non-trivial problem sizes. Existing results analyze specific causes in isolation, such as random deep circuits or global cost functions, but lack a unified picture that integrates circuit design, data encoding, cost structure, and noise. In this work, we develop a comprehensive theory of barren plateaus for parametrized quantum circuits. We prove that three mechanisms jointly control gradient scaling: (1) ansatz-induced concentration, where highly expressive or sufficiently deep circuits approximate unitary 2-designs causing gradient variance to scale as O(2 −n); (2) data-encoding-induced concentration, where certain feature maps wash out class-conditional structure; and (3) probleminduced concentration, where cost functions effectively average over exponentially many degrees of freedom. For each mechanism, we derive explicit scaling laws for the gradient variance as a function of qubit count n, depth L, and encoding complexity. We show that appropriately constrained expressibility, locality, and symmetry breaking can prevent gradients from vanishing exponentially and obtain conditions under which gradients remain at worst polynomially small. This yields architectural and initialization design principles for plateau-resistant circuits, including shallow, problem-informed ansätze, layer-wise training, and structured data encodings. Numerical experiments on up to 100 qubits for VQE, QAOA, and quantum classifiers confirm our theoretical predictions and demonstrate 10-100× improvements in effective gradient magnitudes over standard architectures. Our framework subsumes existing barren plateau results and provides a practical toolkit for diagnosing and mitigating trainability issues in variational quantum algorithms.

•Dec 17, 2025•0 citations

Cost function dependent barren plateaus in shallow parametrized quantum circuits

Marco Cerezo, Akira Sone, Tyler J. Volkoff +2 more

Variational quantum algorithms (VQAs) optimize the parameters θ of a parametrized quantum circuit V(θ) to minimize a cost function C. While VQAs may enable practical applications of noisy quantum computers, they are nevertheless heuristic methods with unproven scaling. Here, we rigorously prove two results, assuming V(θ) is an alternating layered ansatz composed of blocks forming local 2-designs. Our first result states that defining C in terms of global observables leads to exponentially vanishing gradients (i.e., barren plateaus) even when V(θ) is shallow. Hence, several VQAs in the literature must revise their proposed costs. On the other hand, our second result states that defining C with local observables leads to at worst a polynomially vanishing gradient, so long as the depth of V(θ) is [Formula: see text]. Our results establish a connection between locality and trainability. We illustrate these ideas with large-scale simulations, up to 100 qubits, of a quantum autoencoder implementation.

arXiv:2001.00550•Nature Communications•Mar 19, 2021•1,048 citations

Cost-Function-Dependent Barren Plateaus in Shallow Quantum Neural Networks.

Marco Cerezo, Akira Sone, Tyler J. Volkoff +2 more

Variational quantum algorithms (VQAs) optimize the parameters $\boldsymbol{\theta}$ of a quantum neural network $V(\boldsymbol{\theta})$ to minimize a cost function $C$. While VQAs may enable practical applications of noisy quantum computers, they are nevertheless heuristic methods with unproven scaling. Here, we rigorously prove two results, assuming $V(\boldsymbol{\theta})$ is a hardware-efficient ansatz composed of blocks forming local 2-designs. Our first result states that defining $C$ in terms of global observables leads to an exponentially vanishing gradient (i.e., a barren plateau) even when $V(\boldsymbol{\theta})$ is shallow. This implies that several VQAs in the literature must revise their proposed cost functions. On the other hand, our second result states that defining $C$ with local observables leads to at worst a polynomially vanishing gradient, so long as the depth of $V(\boldsymbol{\theta})$ is $\mathcal{O}(\log n)$. Taken together, our results establish a connection between locality and trainability. Finally, we illustrate these ideas with large-scale simulations, up to 100 qubits, of a particular VQA known as quantum autoencoders.

•Jan 2, 2020•82 citations

Tight and Efficient Gradient Bounds for Parameterized Quantum Circuits

Alistair Letcher, Stefan Woerner, Christa Zoufal

The training of a parameterized model largely depends on the landscape of the underlying loss function. In particular, vanishing gradients are a central bottleneck in the scalability of variational quantum algorithms (VQAs), and are known to arise in various ways. However, a caveat of most existing gradient bound results is the requirement of t-design circuit assumptions that are typically not satisfied in practice. In this work, we loosen these assumptions altogether and derive tight upper and lower bounds on loss and gradient concentration for a large class of parameterized quantum circuits and arbitrary observables, which are significantly stronger than prior work. Moreover, we show that these bounds, as well as the variance of the loss itself, can be estimated efficiently and classically-providing practical tools to study the loss landscapes of VQA models, including verifying whether or not a circuit/observable induces barren plateaus. In particular, our results can readily be leveraged to rule out barren plateaus for a realistic class of ansätze and mixed observables, namely, observables containing a non-vanishing local term. This insight has direct implications for hybrid Quantum Generative Adversarial Networks (qGANs). We prove that designing the discriminator appropriately leads to 1-local weights that stay constant in the number of qubits, regardless of discriminator depth. This implies that qGANs with appropriately chosen generators do not suffer from barren plateaus even at scale-making them a promising candidate for applications in generative quantum machine learning. We demonstrate this result by training a qGAN to learn a 2D mixture of Gaussian distributions with up to 16 qubits, and provide numerical evidence that global contributions to the gradient, while initially exponentially small, may kick in substantially over the course of training.

arXiv:2309.12681•Quantum•Sep 25, 2024•28 citations

Can Chemically Inspired Parameter Initialization Mitigate Barren Plateaus in Variational Quantum Eigensolvers?

Zhangyu Yang, Jinzhao Sun, Jianpeng Chen +2 more

Recent advances in the variational quantum eigensolver (VQE) have strengthened its promise for molecular electronic-structure calculations on near-term noisy intermediate-scale quantum (NISQ) devices. However, barren plateaus (BPs) remain a major obstacle to scaling VQE, as vanishing gradients can severely obstruct optimization at large scale. While most BP studies focus on randomly initialized circuits and global gradient behavior, practical quantum chemistry VQE typically uses chemically motivated initializations and local optimization around HF- or MP2-derived states. Although chemically informed initializations are widely expected to improve VQE trainability, how they mitigate local barren plateaus remains unclear. We therefore study the occurrence of local barren plateaus in molecular UCCSD-VQE, considering both chemically motivated initialization points and iterates along the corresponding optimization trajectories. In the benchmarks considered here, chemically initialized starting points and variational trajectories show polynomial decay of local gradient variance with system size instead of the exponential BP-like decay observed for random controls, indicating that chemical initialization mitigates barren plateaus in the local optimization regions and highlights the trainability advantage of chemically motivated VQE strategies.

physics.chem-phCross-listedarXiv:2609.22729•Sep 22, 2026•0 citations
Barren plateaus in variational quantum algorithms are typically described by gradient concentration at random initialization. In contrast, rigorous results for the Hessian, even at the level of entry-wise variance, remain limited. In this work, we analyze the scaling of Hessian-entry variances at initialization. Using exact second-order parameter-shift identities, we write HjkH_{jk}Hjk​ as a constant-size linear combination of shifted cost evaluations, which reduces Varρ(Hjk){\rm Var}_ρ(H_{jk})Varρ​(Hjk​) to a finite-dimensional covariance--quadratic form. For global objectives, under an exponential concentration condition on the cost at initialization, Varρ(Hjk){\rm Var}_ρ(H_{jk})Varρ​(Hjk​) decays exponentially with the number of qubits nnn. For local averaged objectives in bounded-depth circuits, Varρ(Hjk){\rm Var}_ρ(H_{jk})Varρ​(Hjk​) admits polynomial bounds controlled by the growth of the backward lightcone on the interaction graph. As a consequence, the number of measurement shots required to estimate HjkH_{jk}Hjk​ to fixed accuracy inherits the same exponential (global) or polynomial (local) scaling. Extensive numerical experiments over system size, circuit depth, and interaction graphs validate the predicted variance scaling. Overall, the paper quantifies when Hessian entries can be resolved at initialization under finite sampling, providing a mathematically grounded basis for second-order information in variational optimization.
The paper studies whether individual Hessian entries in variational quantum algorithms remain statistically resolvable at random initialization, contrasting global observables with averaged local costs. It considers independently parametrized gates with two-point-spectrum generators, product-uniform initialization, and bounded-depth or sublinearly growing-depth circuits.
Exact second-order parameter-shift identities express each Hessian entry as at most four shifted cost evaluations. Its initialization variance is therefore a constant-dimensional covariance quadratic form. Shift invariance and Cauchy–Schwarz imply that an assumed cost-concentration bound Var⁡(C)≤cge−αn\operatorname{Var}(C)\le c_g e^{-\alpha n}Var(C)≤cg​e−αn transfers to every Hessian entry. For averaged term-wise kkk-local costs, a backward-lightcone dependency-graph argument gives Var⁡(Hjk)≤clocVG(k+2rL)/n\operatorname{Var}(H_{jk})\le c_{\mathrm{loc}}V_G(k+2rL)/nVar(Hjk​)≤cloc​VG​(k+2rL)/n, where VGV_GVG​ measures interaction-graph neighborhood growth. On DDD-dimensional lattices this becomes O((k+2rL)D/n)O((k+2rL)^D/n)O((k+2rL)D/n) at fixed depth.
Statevector simulations, primarily up to 16 qubits with 200 initializations per configuration, report global diagonal/off-diagonal variances approximately e−0.57ne^{-0.57n}e−0.57n and e−0.64ne^{-0.64n}e−0.64n, versus local fits n−1.96n^{-1.96}n−1.96 and n−2.05n^{-2.05}n−2.05. Additional tests examine an Ising-model energy, Hessian spectra, and finite-shot behavior.
Resolution is defined by shot noise falling below a fraction of initialization variance, with a sufficient shot budget proportional to its inverse. The paper interprets this as exponential versus polynomial resolution costs; however, the local theorem supplies only an upper variance bound, not the lower bound needed to guarantee polynomial resolvability. Stated limitations include instance-dependent exponents, conditional global concentration, no spectral-conditioning guarantees, and no device-noise analysis.
Sampling noisy intermediate-scale quantum devices is a fundamental step that converts coherent quantum-circuit outputs to measurement data for running variational quantum algorithms that utilize gradient and Hessian methods in cost-function optimization tasks. This step, however, introduces estimation errors in the resulting gradient or Hessian computations. To minimize these errors, we discuss tunable numerical estimators, which are the finite difference (including their generalized versions) and scaled parameter-shift estimators [introduced in Phys. Rev. A 103, 012405 (2021)], and propose operational circuit-averaged methods to optimize them. We show that these optimized numerical estimators offer estimation errors that drop exponentially with the number of circuit qubits for a given sampling-copy number, revealing a direct compatibility with the barren-plateau phenomenon. In particular, there exists a critical sampling-copy number below which an optimized difference estimator gives a smaller average estimation error in contrast to the standard (analytical) parameter-shift estimator, which exactly computes gradient and Hessian components. Moreover, this critical number grows exponentially with the circuit-qubit number. Finally, by forsaking analyticity, we demonstrate that the scaled parameter-shift estimators beat the standard unscaled ones in estimation accuracy under any situation, with comparable performances to those of the difference estimators within significant copy-number ranges, and are the best ones if larger copy numbers are affordable.
Barren plateaus-regions of exponentially small gradients-are a major obstacle for training variational quantum algorithms on non-trivial problem sizes. Existing results analyze specific causes in isolation, such as random deep circuits or global cost functions, but lack a unified picture that integrates circuit design, data encoding, cost structure, and noise. In this work, we develop a comprehensive theory of barren plateaus for parametrized quantum circuits. We prove that three mechanisms jointly control gradient scaling: (1) ansatz-induced concentration, where highly expressive or sufficiently deep circuits approximate unitary 2-designs causing gradient variance to scale as O(2 −n); (2) data-encoding-induced concentration, where certain feature maps wash out class-conditional structure; and (3) probleminduced concentration, where cost functions effectively average over exponentially many degrees of freedom. For each mechanism, we derive explicit scaling laws for the gradient variance as a function of qubit count n, depth L, and encoding complexity. We show that appropriately constrained expressibility, locality, and symmetry breaking can prevent gradients from vanishing exponentially and obtain conditions under which gradients remain at worst polynomially small. This yields architectural and initialization design principles for plateau-resistant circuits, including shallow, problem-informed ansätze, layer-wise training, and structured data encodings. Numerical experiments on up to 100 qubits for VQE, QAOA, and quantum classifiers confirm our theoretical predictions and demonstrate 10-100× improvements in effective gradient magnitudes over standard architectures. Our framework subsumes existing barren plateau results and provides a practical toolkit for diagnosing and mitigating trainability issues in variational quantum algorithms.
Variational quantum algorithms (VQAs) optimize the parameters θ of a parametrized quantum circuit V(θ) to minimize a cost function C. While VQAs may enable practical applications of noisy quantum computers, they are nevertheless heuristic methods with unproven scaling. Here, we rigorously prove two results, assuming V(θ) is an alternating layered ansatz composed of blocks forming local 2-designs. Our first result states that defining C in terms of global observables leads to exponentially vanishing gradients (i.e., barren plateaus) even when V(θ) is shallow. Hence, several VQAs in the literature must revise their proposed costs. On the other hand, our second result states that defining C with local observables leads to at worst a polynomially vanishing gradient, so long as the depth of V(θ) is [Formula: see text]. Our results establish a connection between locality and trainability. We illustrate these ideas with large-scale simulations, up to 100 qubits, of a quantum autoencoder implementation.
Variational quantum algorithms (VQAs) optimize the parameters θ\boldsymbol{\theta}θ of a quantum neural network V(θ)V(\boldsymbol{\theta})V(θ) to minimize a cost function CCC. While VQAs may enable practical applications of noisy quantum computers, they are nevertheless heuristic methods with unproven scaling. Here, we rigorously prove two results, assuming V(θ)V(\boldsymbol{\theta})V(θ) is a hardware-efficient ansatz composed of blocks forming local 2-designs. Our first result states that defining CCC in terms of global observables leads to an exponentially vanishing gradient (i.e., a barren plateau) even when V(θ)V(\boldsymbol{\theta})V(θ) is shallow. This implies that several VQAs in the literature must revise their proposed cost functions. On the other hand, our second result states that defining CCC with local observables leads to at worst a polynomially vanishing gradient, so long as the depth of V(θ)V(\boldsymbol{\theta})V(θ) is O(log⁡n)\mathcal{O}(\log n)O(logn). Taken together, our results establish a connection between locality and trainability. Finally, we illustrate these ideas with large-scale simulations, up to 100 qubits, of a particular VQA known as quantum autoencoders.
The training of a parameterized model largely depends on the landscape of the underlying loss function. In particular, vanishing gradients are a central bottleneck in the scalability of variational quantum algorithms (VQAs), and are known to arise in various ways. However, a caveat of most existing gradient bound results is the requirement of t-design circuit assumptions that are typically not satisfied in practice. In this work, we loosen these assumptions altogether and derive tight upper and lower bounds on loss and gradient concentration for a large class of parameterized quantum circuits and arbitrary observables, which are significantly stronger than prior work. Moreover, we show that these bounds, as well as the variance of the loss itself, can be estimated efficiently and classically-providing practical tools to study the loss landscapes of VQA models, including verifying whether or not a circuit/observable induces barren plateaus. In particular, our results can readily be leveraged to rule out barren plateaus for a realistic class of ansätze and mixed observables, namely, observables containing a non-vanishing local term. This insight has direct implications for hybrid Quantum Generative Adversarial Networks (qGANs). We prove that designing the discriminator appropriately leads to 1-local weights that stay constant in the number of qubits, regardless of discriminator depth. This implies that qGANs with appropriately chosen generators do not suffer from barren plateaus even at scale-making them a promising candidate for applications in generative quantum machine learning. We demonstrate this result by training a qGAN to learn a 2D mixture of Gaussian distributions with up to 16 qubits, and provide numerical evidence that global contributions to the gradient, while initially exponentially small, may kick in substantially over the course of training.
Recent advances in the variational quantum eigensolver (VQE) have strengthened its promise for molecular electronic-structure calculations on near-term noisy intermediate-scale quantum (NISQ) devices. However, barren plateaus (BPs) remain a major obstacle to scaling VQE, as vanishing gradients can severely obstruct optimization at large scale. While most BP studies focus on randomly initialized circuits and global gradient behavior, practical quantum chemistry VQE typically uses chemically motivated initializations and local optimization around HF- or MP2-derived states. Although chemically informed initializations are widely expected to improve VQE trainability, how they mitigate local barren plateaus remains unclear. We therefore study the occurrence of local barren plateaus in molecular UCCSD-VQE, considering both chemically motivated initialization points and iterates along the corresponding optimization trajectories. In the benchmarks considered here, chemically initialized starting points and variational trajectories show polynomial decay of local gradient variance with system size instead of the exponential BP-like decay observed for random controls, indicating that chemical initialization mitigates barren plateaus in the local optimization regions and highlights the trainability advantage of chemically motivated VQE strategies.