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:2611.00272

Triple descent in quantum linear regression

Gabriele Lo Monaco, Salvatore Lorenzo, Alessandro Ferraro, Mauro Paternostro, G. Massimo Palma, Luca Innocenti

quant-phSubmitted Oct 30, 2026• Announced Nov 3, 2026• 0 citations
arXivHTMLPDF

Abstract

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.

AI Summary

Quantum linear regression learns linear readouts of measurement-derived features, which are estimated from finite shots. This paper asks how sampling changes generalization as the number of measurement outcomes $n_{\mathrm{out}}$ varies, separating observable-reconstruction error from test-shot variance: $\mathrm{MSE}=\mathcal E_N+V_N/M$, with $N$ training and $M$ test shots per state. For $n_{\mathrm{tr}}>d^2$, it identifies two peaks producing triple descent. A classical peak at informational completeness, $n_{\mathrm{out}}=d^2$, comes from ill-conditioning of the POVM frame. For Haar-random rank-one POVMs, its rescaled spectrum converges to Marchenko–Pastur, yielding normalized exact-reconstruction variance $\min\{1,\gamma_C^{-1}\}/|1-\gamma_C|$, where $\gamma_C=d^2/n_{\mathrm{out}}$. Finite training noise can suppress this peak. A second peak at $n_{\mathrm{out}}=n_{\mathrm{tr}}>d^2$ arises because shot noise fills feature directions absent from exact Born probabilities. Under bounded-frame, nondegenerate-noise and label-variation assumptions, ridgeless test variance diverges in probability in proportional high-dimensional limits. Near interpolation, ridge-rounded singular profiles scale as $[(\gamma_Q-1)^2+4h\eta]^{-1/2}$, with $\gamma_Q=n_{\mathrm{out}}/n_{\mathrm{tr}}$ and $\eta=n_{\mathrm{out}}N\tau$. Random-matrix analysis of multinomial records and simulations establish these results. Increasing training shots reduces reconstruction error but leaves a nonzero variance-peak amplitude; at fixed dimension, variance saturation is argued heuristically. Exact training features remove this second peak. Limitations include geometry and scaling assumptions, unresolved finite-size peak distributions, and unproved fixed-dimensional convergence. Extensions to nonlinear readouts and fidelity kernels remain open.

Little sign of AI-generated textOverclaims

Error found: The abstract omits that the demonstrated triple-descent profile requires suitable measurement geometry, an essentially unregularized estimator, and n_tr > d²; symmetric measurements can avoid the classical peak. The abstract's statement that the classical peak affects predictions only through variance requires exact target reconstruction and does not generally apply to finite-shot training, which can also introduce observable-reconstruction bias along weak measurement directions.

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

Comments
Main: 20 pages, 12 figures. Supplementary: 50 pages

Related papers

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

Shaping double descent in quantum extreme learning through model design

Sabri Meyer, Ege Yilmaz, Aurelien Lucchi +2 more

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.

quant-pharXiv:2610.06581•Oct 6, 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

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

On fundamental aspects of quantum extreme learning machines

Weijie Xiong, Giorgio Facelli, Mehrad Sahebi +4 more

Abstract Quantum extreme learning machines (QELMs) have emerged as a promising framework for quantum machine learning. Their appeal lies in the rich feature map induced by the dynamics of a quantum substrate—the quantum reservoir—and the efficient post-measurement training via linear regression. Here, we study the expressivity of QELMs by decomposing the prediction of QELMs into a Fourier series. We show that the achievable Fourier frequencies are determined by the data encoding scheme, while Fourier coefficients depend on both the reservoir and the measurement. Notably, the expressivity of QELMs is fundamentally limited by the number of Fourier frequencies and the number of observables, while the complexity of the prediction hinges on the reservoir. As a cautionary note on scalability, we identify four sources that can lead to the exponential concentration of the observables as the system size grows (randomness, hardware noise, entanglement, and global measurements) and show how this can turn QELMs into useless input-agnostic oracles. In particular, our result on the reservoir-induced concentration strongly indicates that quantum reservoirs drawn from a highly random ensemble make QELM models unscalable. Our analysis elucidates the potential and fundamental limitations of QELMs and lays the groundwork for systematically exploring quantum reservoir systems for other machine learning tasks.

arXiv:2312.15124•Quantum Machine Intelligence•Feb 13, 2025•75 citations
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 linear regression learns linear readouts of measurement-derived features, which are estimated from finite shots. This paper asks how sampling changes generalization as the number of measurement outcomes noutn_{\mathrm{out}}nout​ varies, separating observable-reconstruction error from test-shot variance: MSE=EN+VN/M\mathrm{MSE}=\mathcal E_N+V_N/MMSE=EN​+VN​/M, with NNN training and MMM test shots per state.
For ntr>d2n_{\mathrm{tr}}>d^2ntr​>d2, it identifies two peaks producing triple descent. A classical peak at informational completeness, nout=d2n_{\mathrm{out}}=d^2nout​=d2, comes from ill-conditioning of the POVM frame. For Haar-random rank-one POVMs, its rescaled spectrum converges to Marchenko–Pastur, yielding normalized exact-reconstruction variance min⁡{1,γC−1}/∣1−γC∣\min\{1,\gamma_C^{-1}\}/|1-\gamma_C|min{1,γC−1​}/∣1−γC​∣, where γC=d2/nout\gamma_C=d^2/n_{\mathrm{out}}γC​=d2/nout​. Finite training noise can suppress this peak. A second peak at nout=ntr>d2n_{\mathrm{out}}=n_{\mathrm{tr}}>d^2nout​=ntr​>d2 arises because shot noise fills feature directions absent from exact Born probabilities. Under bounded-frame, nondegenerate-noise and label-variation assumptions, ridgeless test variance diverges in probability in proportional high-dimensional limits. Near interpolation, ridge-rounded singular profiles scale as [(γQ−1)2+4hη]−1/2[(\gamma_Q-1)^2+4h\eta]^{-1/2}[(γQ​−1)2+4hη]−1/2, with γQ=nout/ntr\gamma_Q=n_{\mathrm{out}}/n_{\mathrm{tr}}γQ​=nout​/ntr​ and η=noutNτ\eta=n_{\mathrm{out}}N\tauη=nout​Nτ.
Random-matrix analysis of multinomial records and simulations establish these results. Increasing training shots reduces reconstruction error but leaves a nonzero variance-peak amplitude; at fixed dimension, variance saturation is argued heuristically. Exact training features remove this second peak. Limitations include geometry and scaling assumptions, unresolved finite-size peak distributions, and unproved fixed-dimensional convergence. Extensions to nonlinear readouts and fidelity kernels remain open.
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.
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 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.
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.
Abstract Quantum extreme learning machines (QELMs) have emerged as a promising framework for quantum machine learning. Their appeal lies in the rich feature map induced by the dynamics of a quantum substrate—the quantum reservoir—and the efficient post-measurement training via linear regression. Here, we study the expressivity of QELMs by decomposing the prediction of QELMs into a Fourier series. We show that the achievable Fourier frequencies are determined by the data encoding scheme, while Fourier coefficients depend on both the reservoir and the measurement. Notably, the expressivity of QELMs is fundamentally limited by the number of Fourier frequencies and the number of observables, while the complexity of the prediction hinges on the reservoir. As a cautionary note on scalability, we identify four sources that can lead to the exponential concentration of the observables as the system size grows (randomness, hardware noise, entanglement, and global measurements) and show how this can turn QELMs into useless input-agnostic oracles. In particular, our result on the reservoir-induced concentration strongly indicates that quantum reservoirs drawn from a highly random ensemble make QELM models unscalable. Our analysis elucidates the potential and fundamental limitations of QELMs and lays the groundwork for systematically exploring quantum reservoir systems for other machine learning tasks.