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:2610.06581

Shaping double descent in quantum extreme learning through model design

Sabri Meyer, Ege Yilmaz, Aurelien Lucchi, Francesco Tacchino, Francesco Scala

quant-phSubmitted Oct 5, 2026• Announced Oct 6, 2026• 1 citation
arXivHTMLPDF

Abstract

Increasing capacity can induce a machine learning model to generalize worse until it perfectly interpolates the training data, and better once it moves beyond interpolation. This inversion, known as \emph{double descent}, is well understood for independent features. For correlated features, its shape is governed by the spectrum of the feature covariance, but existing analyses typically yield only implicit equations. We first close this gap for classical linear regression, analytically characterizing the entire risk curve through two quantities, namely the condition number of the feature population covariance and a spectral measure of correlation strength. Then, we apply this theory on quantum models, focusing on quantum extreme learning machines (QELMs). We show that three tunable design knobs, \textit{i.e.,} encoding strength, input reuploads, and measurement shots, each regularize the QELM output covariance, and derive conditions under which the interpolation threshold remains stable. Finite measurement shot noise can be leveraged to guarantee this stability. Numerical experiments confirm each prediction, yielding concrete design principles for QELMs that generalize well beyond interpolation.

AI Summary

The paper asks how circuit architecture and measurement noise shape double descent in quantum extreme learning machines (QELMs), which fit a minimum-norm classical linear readout to fixed quantum features. It connects linear-regression risk bounds to the population second-moment matrix $\Sigma=\mathbb{E}[\phi(x)\phi(x)^\top]$, then uses the features’ finite Fourier expansion to analyze encoding strength, input reuploads, and measurement shots. For underparameterized, realizable regression, it recovers $\mathcal{R}(\gamma)=\sigma^2/(1-\gamma)$, with $\gamma=d/n$, under the sufficient correlation condition $\lambda_{\max}(\operatorname{Cov}(z\otimes z))=o(d)$, where $z=\Sigma^{-1/2}\phi(x)$. For large $\gamma$, sub-Gaussian whitened features satisfying $\|z\|=\sqrt d$ almost surely give $\mathcal{R}(\gamma)=\sigma^2+\mathcal{O}(\|w_*\|^2\lambda_{\max}(\Sigma)+\sigma^2\kappa(\Sigma)/(\sqrt\gamma-C_K)^2)$. These results use generalized Marchenko–Pastur theory and singular-value concentration. Strong input scaling approaches a Fourier-coefficient covariance, with conditioning corrections $\mathcal{O}(s^{-K})$ under density regularity assumptions. Additional reuploads improve a conditioning bound when new frequencies populate the weakest covariance eigenspace. Independent measurement batches add diagonal shot-noise covariance; under stated observable assumptions, $m_0=o(\sqrt d)$ shots per feature suffice for the correlation condition preserving the canonical threshold. Statevector experiments up to 16 qubits and real-data tests examine these predictions. The stated limitations include restrictive geometric assumptions for the overparameterized bound, incomplete justification for non-Gaussian inputs and moderate dimensions, and no general guarantee for the reupload injection condition. The analysis does not explain secondary risk peaks or guarantee that overparameterization outperforms the underparameterized regime.

Little sign of AI-generated textOverclaimsIncrementalStrong assumptionsPossible error

Central claim may not hold: The advertised characterization of the entire double-descent curve is substantially stronger than the paper's conditional, large-overparameterization upper bound. The claimed universal small-encoding condition-number divergence fails for valid quantum features (sin(s x_1), sin(s x_2)) whose second-moment matrix remains proportional to the identity. The reupload bound requires architecture-dependent injection assumptions and does not generally compare the deeper circuit's conditioning with that of the original circuit. For labels generated from exact quantum features, finite-shot features produce regressor-correlated residuals, invalidating direct application of the paper's independent-label-noise risk theorem. The assumptions used to obtain Marchenko–Pastur convergence do not ensure convergence of inverse spectral traces or expected risk, and permit counterexamples to the stated under-parametrized risk formula.

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

Comments
Main: 10 pages, 2 figures; Supplementary: 89 pages, 5 figures

Related papers

Triple descent in quantum linear regression

Gabriele Lo Monaco, Salvatore Lorenzo, Alessandro Ferraro +3 more

Quantum linear regression models, such as quantum extreme learning machines and quantum fidelity kernels, are trained on features that can only be estimated from finitely many measurement shots. Sampling noise is usually treated as a nuisance that more shots suppress; we show instead that it can fundamentally reshape the instabilities of the learning problem. As the number of measurement outcomes grows, the prediction error displays a triple-descent profile, with two peaks of fundamentally different origins. The first, classical, peak occurs when the measurement first becomes informationally complete, is a direct analogue of double descent in classical linear regression, and is caused by ill-conditioning of the measurement itself. This peak affects predictions only through their variance, and is thus suppressed by increasing the number of shots at inference. The second, quantum, peak occurs beyond informational completeness when the number of outcomes matches the number of training states: shot noise populates unphysical outcome combinations and the regression fits these pure-noise features with large weights, corrupting the learned model itself. More training or test shots cannot remove this quantum peak, which disappears only in the ideal limit of exact training features. Using random-matrix theory, we derive the shape of both peaks analytically, accounting also for ridge regularization. Our results show that treating quantum features as exactly known overestimates classical instabilities while missing genuinely quantum ones.

quant-pharXiv:2611.00272•Nov 3, 2026•0 citations

Who can sample forever? Shot noise effects in quantum linear regression

Gabriele Lo Monaco, Salvatore Lorenzo, Alessandro Ferraro +3 more

Quantum fidelity kernels and quantum extreme learning machines share the same basic architecture: a fixed quantum device generates features that are combined through a classically trained linear model. We refer to this broad non-variational setting as quantum linear regression. A defining aspect of these models is that the quantum features are not directly available, but must be estimated from a finite number of shots. We show that finite statistics makes quantum linear regression unavoidably ill-conditioned. Ideal feature vectors are confined to a subspace determined by the dimension of the Hilbert space, whereas finite-shot estimates generically acquire components outside this subspace. Training then attempts to fit directions that disappear in the infinite-statistics limit, making the learned model increasingly sensitive to sampling noise. We analyze this mechanism for quantum extreme learning machines trained on $n_{\rm tr}$ states with $N$ shots per state. Despite the ill-conditioning, predictions remain controlled when the training targets are exact. Finite sampling produces a systematic prediction error that decreases quadratically with $N$, while fluctuations induced by the random training measurements decrease inversely with the total training budget. Thus, unlike in standard linear regression with exactly known features, increasing the dataset size alone cannot remove the error caused by sampling noise. If the training targets are also noisy, due for instance to state-preparation errors, the otherwise hidden directions become active. The test error can then increase with $N$, a behavior that we ascribe to a form of quantum overfitting. Our results expose a fundamental tradeoff between dataset size and per-state statistics in non-variational quantum learning and provide a baseline for resource-aware training and regularization strategies.

quant-pharXiv:2610.12015•Oct 9, 2026•1 citation

Double Descent in Quantum Kernel Ridge Regression

Kensuke Kamisoyama, Lento Nagano, Koji Terashi

Various classical machine learning models, including linear regression, kernel methods, and deep neural networks, exhibit double descent, in which the test risk peaks near the interpolation threshold and then decreases in the overparameterized regime. However, this phenomenon has received less attention in the quantum setting. In this work, we investigate the double descent phenomenon in quantum kernel ridge regression (QKRR). By applying deterministic equivalents from random matrix theory (RMT), we derive an asymptotic expression for the test risk of QKRR in the high-dimensional limit. Our analysis rigorously characterizes the interpolation peak and reveals how explicit regularization can effectively suppress it. We corroborate our theoretical results with numerical simulations, demonstrating close agreement even for finite-size quantum systems.

quant-pharXiv:2604.17202•Apr 21, 2026•4 citations

Quantum inference on a classically trained quantum extreme learning machine

Emanuele Brusaschi, Marco Clementi, Marco Liscidini +3 more

Quantum extreme learning machines (QELMs) are unconventional computing architectures that bear remarkable promise in both classical and quantum machine-learning tasks, such as the estimation of quantum state properties. However, the probabilistic nature of quantum measurements demands extensive repetitions for training to precisely estimate expectation values, imposing stringent trade-offs among experimental resources, acquisition time, and signal-to-noise ratio, particularly for large datasets. Here, we introduce a paradigm shift by training the QELM exclusively with intense classical fields, namely twin beams generated by stimulated emission, while performing inference directly on previously unseen genuine quantum input states to predict their quantum properties. This strategy dramatically reduces acquisition times while substantially enhancing the signal-to-noise ratio. Using frequency-bin encoded biphoton states, implemented here for the first time in a quantum machine learning architecture, we demonstrate entanglement witnessing of two-qubit states with (93 +- 4)% accuracy, multi-dimensional entanglement detection, and learning of the Hamiltonian governing photon-pair generation with a fidelity of (96 +- 4)%. Our results open a new pathway toward faster and more robust training of photonic quantum extreme learning machines for quantum feature extraction.

quant-pharXiv:2603.20167•Mar 23, 2026•1 citation

Double Descent in Quantum Kernel Methods

Marie Kempkes, Aroosa Ijaz, Elies Gil-Fuster +4 more

The double descent phenomenon challenges traditional statistical learning theory by revealing scenarios where larger models do not necessarily lead to reduced performance on unseen data. While this counterintuitive behavior has been observed in a variety of classical machine learning models, particularly modern neural network architectures, it remains elusive within the context of quantum machine learning. In this work, we analytically demonstrate that linear regression models in quantum feature spaces can exhibit double descent behavior by drawing on insights from classical linear regression and random matrix theory. Additionally, our numerical experiments on quantum kernel methods across different real-world datasets and system sizes further confirm the existence of a test error peak, a characteristic feature of double descent. Our findings provide evidence that quantum models can operate in the modern, overparameterized regime without experiencing overfitting, potentially opening pathways to improved learning performance beyond traditional statistical learning theory.

arXiv:2501.10077•PRX Quantum•Sep 15, 2025•16 citations

Cautious optimism for deep parameterized quantum circuits

Marie Kempkes, Elies Gil-Fuster, Carlos Bravo-Prieto +5 more

A central challenge in quantum machine learning is understanding the scaling behavior of parameterized quantum circuits (PQCs). In particular, it remains unclear how their performance on unseen data changes as the number of trainable parameters increases. Prior works have derived formal generalization guarantees for quantum models, but it is well-known that many such results do not fully characterize generalization behavior in practice. In this work, we show that gradient-based PQCs can exhibit improved performance on unseen data as model size increases, displaying the phenomenon of double descent. This contrasts with the traditional view that larger models lead to degraded generalization. We provide analytical results rigorously underpinning this behavior by leveraging add-one-in perturbation techniques and spectral properties of random matrices. We support these results with numerical experiments on re-uploading PQCs across several data sets and training set sizes, consistently observing the predicted double descent behavior. While other obstacles on the path toward practical quantum machine learning remain, our finding that deeper parameterized quantum circuits do not necessarily exhibit degraded performance provides reasons for cautious optimism.

quant-pharXiv:2607.21409•Jul 24, 2026•1 citation
Increasing capacity can induce a machine learning model to generalize worse until it perfectly interpolates the training data, and better once it moves beyond interpolation. This inversion, known as \emph{double descent}, is well understood for independent features. For correlated features, its shape is governed by the spectrum of the feature covariance, but existing analyses typically yield only implicit equations. We first close this gap for classical linear regression, analytically characterizing the entire risk curve through two quantities, namely the condition number of the feature population covariance and a spectral measure of correlation strength. Then, we apply this theory on quantum models, focusing on quantum extreme learning machines (QELMs). We show that three tunable design knobs, \textit{i.e.,} encoding strength, input reuploads, and measurement shots, each regularize the QELM output covariance, and derive conditions under which the interpolation threshold remains stable. Finite measurement shot noise can be leveraged to guarantee this stability. Numerical experiments confirm each prediction, yielding concrete design principles for QELMs that generalize well beyond interpolation.
The paper asks how circuit architecture and measurement noise shape double descent in quantum extreme learning machines (QELMs), which fit a minimum-norm classical linear readout to fixed quantum features. It connects linear-regression risk bounds to the population second-moment matrix Σ=E[ϕ(x)ϕ(x)⊤]\Sigma=\mathbb{E}[\phi(x)\phi(x)^\top]Σ=E[ϕ(x)ϕ(x)⊤], then uses the features’ finite Fourier expansion to analyze encoding strength, input reuploads, and measurement shots.
For underparameterized, realizable regression, it recovers R(γ)=σ2/(1−γ)\mathcal{R}(\gamma)=\sigma^2/(1-\gamma)R(γ)=σ2/(1−γ), with γ=d/n\gamma=d/nγ=d/n, under the sufficient correlation condition λmax⁡(Cov⁡(z⊗z))=o(d)\lambda_{\max}(\operatorname{Cov}(z\otimes z))=o(d)λmax​(Cov(z⊗z))=o(d), where z=Σ−1/2ϕ(x)z=\Sigma^{-1/2}\phi(x)z=Σ−1/2ϕ(x). For large γ\gammaγ, sub-Gaussian whitened features satisfying ∥z∥=d\|z\|=\sqrt d∥z∥=d​ almost surely give R(γ)=σ2+O(∥w∗∥2λmax⁡(Σ)+σ2κ(Σ)/(γ−CK)2)\mathcal{R}(\gamma)=\sigma^2+\mathcal{O}(\|w_*\|^2\lambda_{\max}(\Sigma)+\sigma^2\kappa(\Sigma)/(\sqrt\gamma-C_K)^2)R(γ)=σ2+O(∥w∗​∥2λmax​(Σ)+σ2κ(Σ)/(γ​−CK​)2). These results use generalized Marchenko–Pastur theory and singular-value concentration.
Strong input scaling approaches a Fourier-coefficient covariance, with conditioning corrections O(s−K)\mathcal{O}(s^{-K})O(s−K) under density regularity assumptions. Additional reuploads improve a conditioning bound when new frequencies populate the weakest covariance eigenspace. Independent measurement batches add diagonal shot-noise covariance; under stated observable assumptions, m0=o(d)m_0=o(\sqrt d)m0​=o(d​) shots per feature suffice for the correlation condition preserving the canonical threshold. Statevector experiments up to 16 qubits and real-data tests examine these predictions.
The stated limitations include restrictive geometric assumptions for the overparameterized bound, incomplete justification for non-Gaussian inputs and moderate dimensions, and no general guarantee for the reupload injection condition. The analysis does not explain secondary risk peaks or guarantee that overparameterization outperforms the underparameterized regime.
Quantum linear regression models, such as quantum extreme learning machines and quantum fidelity kernels, are trained on features that can only be estimated from finitely many measurement shots. Sampling noise is usually treated as a nuisance that more shots suppress; we show instead that it can fundamentally reshape the instabilities of the learning problem. As the number of measurement outcomes grows, the prediction error displays a triple-descent profile, with two peaks of fundamentally different origins. The first, classical, peak occurs when the measurement first becomes informationally complete, is a direct analogue of double descent in classical linear regression, and is caused by ill-conditioning of the measurement itself. This peak affects predictions only through their variance, and is thus suppressed by increasing the number of shots at inference. The second, quantum, peak occurs beyond informational completeness when the number of outcomes matches the number of training states: shot noise populates unphysical outcome combinations and the regression fits these pure-noise features with large weights, corrupting the learned model itself. More training or test shots cannot remove this quantum peak, which disappears only in the ideal limit of exact training features. Using random-matrix theory, we derive the shape of both peaks analytically, accounting also for ridge regularization. Our results show that treating quantum features as exactly known overestimates classical instabilities while missing genuinely quantum ones.
Quantum fidelity kernels and quantum extreme learning machines share the same basic architecture: a fixed quantum device generates features that are combined through a classically trained linear model. We refer to this broad non-variational setting as quantum linear regression. A defining aspect of these models is that the quantum features are not directly available, but must be estimated from a finite number of shots. We show that finite statistics makes quantum linear regression unavoidably ill-conditioned. Ideal feature vectors are confined to a subspace determined by the dimension of the Hilbert space, whereas finite-shot estimates generically acquire components outside this subspace. Training then attempts to fit directions that disappear in the infinite-statistics limit, making the learned model increasingly sensitive to sampling noise. We analyze this mechanism for quantum extreme learning machines trained on ntrn_{\rm tr}ntr​ states with NNN shots per state. Despite the ill-conditioning, predictions remain controlled when the training targets are exact. Finite sampling produces a systematic prediction error that decreases quadratically with NNN, while fluctuations induced by the random training measurements decrease inversely with the total training budget. Thus, unlike in standard linear regression with exactly known features, increasing the dataset size alone cannot remove the error caused by sampling noise. If the training targets are also noisy, due for instance to state-preparation errors, the otherwise hidden directions become active. The test error can then increase with NNN, a behavior that we ascribe to a form of quantum overfitting. Our results expose a fundamental tradeoff between dataset size and per-state statistics in non-variational quantum learning and provide a baseline for resource-aware training and regularization strategies.
Various classical machine learning models, including linear regression, kernel methods, and deep neural networks, exhibit double descent, in which the test risk peaks near the interpolation threshold and then decreases in the overparameterized regime. However, this phenomenon has received less attention in the quantum setting. In this work, we investigate the double descent phenomenon in quantum kernel ridge regression (QKRR). By applying deterministic equivalents from random matrix theory (RMT), we derive an asymptotic expression for the test risk of QKRR in the high-dimensional limit. Our analysis rigorously characterizes the interpolation peak and reveals how explicit regularization can effectively suppress it. We corroborate our theoretical results with numerical simulations, demonstrating close agreement even for finite-size quantum systems.
Quantum extreme learning machines (QELMs) are unconventional computing architectures that bear remarkable promise in both classical and quantum machine-learning tasks, such as the estimation of quantum state properties. However, the probabilistic nature of quantum measurements demands extensive repetitions for training to precisely estimate expectation values, imposing stringent trade-offs among experimental resources, acquisition time, and signal-to-noise ratio, particularly for large datasets. Here, we introduce a paradigm shift by training the QELM exclusively with intense classical fields, namely twin beams generated by stimulated emission, while performing inference directly on previously unseen genuine quantum input states to predict their quantum properties. This strategy dramatically reduces acquisition times while substantially enhancing the signal-to-noise ratio. Using frequency-bin encoded biphoton states, implemented here for the first time in a quantum machine learning architecture, we demonstrate entanglement witnessing of two-qubit states with (93 +- 4)% accuracy, multi-dimensional entanglement detection, and learning of the Hamiltonian governing photon-pair generation with a fidelity of (96 +- 4)%. Our results open a new pathway toward faster and more robust training of photonic quantum extreme learning machines for quantum feature extraction.
The double descent phenomenon challenges traditional statistical learning theory by revealing scenarios where larger models do not necessarily lead to reduced performance on unseen data. While this counterintuitive behavior has been observed in a variety of classical machine learning models, particularly modern neural network architectures, it remains elusive within the context of quantum machine learning. In this work, we analytically demonstrate that linear regression models in quantum feature spaces can exhibit double descent behavior by drawing on insights from classical linear regression and random matrix theory. Additionally, our numerical experiments on quantum kernel methods across different real-world datasets and system sizes further confirm the existence of a test error peak, a characteristic feature of double descent. Our findings provide evidence that quantum models can operate in the modern, overparameterized regime without experiencing overfitting, potentially opening pathways to improved learning performance beyond traditional statistical learning theory.
A central challenge in quantum machine learning is understanding the scaling behavior of parameterized quantum circuits (PQCs). In particular, it remains unclear how their performance on unseen data changes as the number of trainable parameters increases. Prior works have derived formal generalization guarantees for quantum models, but it is well-known that many such results do not fully characterize generalization behavior in practice. In this work, we show that gradient-based PQCs can exhibit improved performance on unseen data as model size increases, displaying the phenomenon of double descent. This contrasts with the traditional view that larger models lead to degraded generalization. We provide analytical results rigorously underpinning this behavior by leveraging add-one-in perturbation techniques and spectral properties of random matrices. We support these results with numerical experiments on re-uploading PQCs across several data sets and training set sizes, consistently observing the predicted double descent behavior. While other obstacles on the path toward practical quantum machine learning remain, our finding that deeper parameterized quantum circuits do not necessarily exhibit degraded performance provides reasons for cautious optimism.