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

Journal of Quantum Artificial Intelligence

Self-Supervised Contrastive Learning in Quantum Latent Spaces for Few-Shot Domain Adaptation

Christianto Hernando

Published Oct 17, 2026• 0 citations
PDF (CC BY)Publisher

Abstract

Few-shot domain adaptation remains fragile under compound distribution shifts because sparse target supervision amplifies representation noise and destabilizes metric structure. This study introduces self-supervised contrastive learning directly in a quantum latent space produced by a variational quantum circuit encoder, followed by prototype-driven adaptation with distribution alignment. Across three target shifts (style, temporal sparsity, and mixed covariate plus conditional drift), the proposed method achieved mean episodic accuracies of 0.721, 0.705, and 0.689 at 1-shot, increasing to 0.832, 0.813, and 0.806 at 5-shot, and 0.851, 0.836, and 0.826 at 10-shot, respectively. At K=5, average accuracy improved from 0.672 under ERM to 0.824 with the full quantum-contrastive model, with 95% confidence intervals tightening from ±0.018 to ±0.011. Representation diagnostics showed a reduction in intra-class dispersion from 0.412 to 0.301 and an increase in inter-class separation from 0.688 to 0.824, while seed variability decreased from 0.012 to 0.007. Circuit ablations identified an optimal depth range of 6–8 layers, where accuracy peaked at 0.824–0.829 before declining at higher depth with increased episodic variance. Robustness tests under finite-shot measurement revealed graceful degradation: under severe shift, accuracy rose from 0.709 at 64 shots per observable to 0.801 at 2048 shots, with failure episodes reduced from 4.2% to 2.9% when increasing from 256 to 1024 shots. These results indicate that quantum-measurement-bounded embeddings can stabilize contrastive geometry and improve few-shot transfer when target labels are scarce.

DOI
10.63913/jqai.v1i3.13

Related papers

Supervised Latent Restructuring for Fixed and Adaptive Quantum Kernel Learning in Small-Data Image Classification

Alakananda Mitra, David H. Fleisher, Chittaranjan Ray +1 more

Fine-grained image classification under limited labeled data presents a representation challenge: pretrained deep networks provide rich but high-dimensional features, whereas small-data learning and practical quantum encoding require substantially more compact representations. Direct dimensionality reduction can make such features computationally tractable but may not preserve the class-discriminative geometry needed for effective kernel learning. This study, therefore, examines whether supervised latent restructuring can produce a compact, class-aware, qubit-compatible representation that supports classical and quantum-kernel learning. EfficientNet-B0 embeddings were reduced from 1280 to 64 dimensions using PCA and subsequently restructured by LDA into an 11-dimensional Supervised Latent Restructuring (SLR) representation. SLR increased the held-out Silhouette coefficient from −0.0054 in PCA64 to 0.0490; however, predictive effects were model dependent: Random Forest and XGBoost improved, Linear SVM changed little, and RBF-SVM favored PCA64. A dimension-matched PCA11 control showed no consistent predictive advantage of SLR over PCA11 across the evaluated classifiers. Under the original 12-class taxonomy, the fixed quantum kernel remained below matched classical baselines. Under a broader three-group taxonomy, quantum-kernel performance improved with increasing training support, but RBF-SVM remained the strongest. $\gamma $ -only Quantum Kernel Alignment did not consistently outperform the fixed kernel and incurred up to $420\times $ greater runtime than the fixed kernel. Target-control and kernel-structure analyses showed that increased alignment produced little additional class-discriminative structure. Overall, the study shows that representation geometry, label granularity, numerical scaling, and adaptive feature scaling jointly influence small-data quantum-kernel learning; no quantum advantage was demonstrated.

IEEE Access•Jan 1, 2026•0 citations

Uncertainty-Calibrated Quantum Neural Networks via Bayesian Variational Circuits and Conformal Prediction

Bryan Daniel Angelino

Reliable decision-making with quantum neural networks on NISQ hardware requires uncertainty estimates that remain meaningful under finite-shot measurement noise and distribution shift. This study introduces an end-to-end pipeline that combines Bayesian variational quantum circuits with post hoc probability calibration and conformal prediction to produce coverage-controlled prediction sets. Across shot budgets from 128 to 4096 shots, the Bayesian variational circuit improves accuracy from 0.828 to 0.839 at 128 shots and from 0.862 to 0.892 at 4096 shots, while reducing negative log-likelihood from 0.462 to 0.448 at 128 shots and from 0.392 to 0.357 at 4096 shots. Calibration preserves accuracy while further lowering negative log-likelihood to 0.432 at 128 shots and 0.350 at 4096 shots, and reduces expected calibration error from 0.058 to 0.027. Conformalization achieves empirical coverage of 0.952 at miscoverage 0.05, with average set size 1.39 and singleton rate 0.71, and maintains coverage between 0.944 and 0.948 under feature corruption and subpopulation shift, where deterministic accuracy degrades to 0.801 and 0.774, respectively, while the Bayesian model retains 0.842 and 0.816. Ablations identify an operational knee at circuit depth 8, and show diminishing returns beyond 8 posterior samples per prediction, supporting deployment-efficient uncertainty control for quantum learning.

Journal of Quantum Artificial Intelligence•Aug 31, 2026•0 citations

Quantum Machine Learning via Contrastive Training

Liudmila A. Zhukas, Vivian Ni Zhang, Qiang Miao +5 more

Quantum machine learning (QML) has attracted growing interest with the rapid parallel advances in large-scale classical machine learning and quantum technologies. Similar to classical machine learning, QML models also face challenges arising from the scarcity of labeled data, particularly as their scale and complexity increase. Here, we introduce self-supervised pretraining of quantum representations that reduces reliance on labeled data by learning invariances from unlabeled examples. We implement this paradigm on a programmable trapped-ion quantum computer, encoding images as quantum states. In situ contrastive pretraining on hardware yields a representation that, when fine-tuned, classifies image families with higher mean test accuracy and lower run-to-run variability than models trained from random initialization. Performance improvement is especially significant in regimes with limited labeled training data. We show that the learned invariances generalize beyond the pretraining image samples. Unlike prior work, our pipeline derives similarity from measured quantum overlaps and executes all training and classification stages on hardware. These results establish a label-efficient route to quantum representation learning, with direct relevance to quantum-native datasets and a clear path to larger classical inputs.

cs.LGCross-listedarXiv:2511.13497•Nov 18, 2025•1 citation

Hybrid Classical–Quantum Transfer Learning Using Variational Quantum Layers for Low-Data Regimes

Krishna Munishamaiah, Krishna Munishamaiah

Supplementary Data Description This supplementary dataset provides the complete experimental results underlying Figures 4–9 of the manuscript. Each CSV file corresponds to one figure and reports mean performance values (and standard deviations where applicable), computed over multiple independent runs with different random seeds. Fig4_Low_Data_Performance.csv Contains classification accuracy (mean ± standard deviation) for Linear, MLP, and VQL heads across varying labeled target data fractions (1%, 5%, 10%, 20%). This dataset supports the low-data performance analysis presented in Fig. 4. Fig5_Parameter_Matched.csv Reports accuracy statistics for parameter-matched adaptation heads, including Linear Probe, MLP Adapter, Bottleneck Adapter, and the proposed VQL Head. This file enables fair comparison of adaptation strategies under comparable trainable parameter budgets (Fig. 5). Fig6_Domain_Shift.csv Provides source-domain accuracy, target-domain accuracy, and the resulting accuracy drop for each model under domain shift conditions. These results quantify robustness to distributional changes, as shown in Fig. 6. Fig7_Ablation.csv Contains ablation results evaluating the effect of quantum circuit depth and number of qubits on classification accuracy. This dataset supports the architectural analysis in Fig. 7. Fig8_Noise_Aware.csv Includes classification accuracy values under increasing quantum noise probabilities for all adaptation heads, enabling noise robustness evaluation corresponding to Fig. 8. Fig9_Convergence.csv Reports training accuracy as a function of training epochs for Linear, MLP, and VQL heads, illustrating convergence behavior and optimization stability (Fig. 9). Reproducibility Note All values are reported as mean performance over multiple independent runs, ensuring statistical reliability and reproducibility of the experimental findings.

•Jan 6, 2026•0 citations

Variational Quantum Algorithms for Image Classification

Andrej Sum-Shik, Peter Thoma

Abstract Variational quantum algorithms are among the most widely studied approaches for applying near-term quantum devices to learning tasks. In this work we investigate a supervised quantum-autoencoder-style classifier that encodes images by amplitude encoding and predicts labels from a designated readout sub-register. Building on the quantum-autoencoder (QAE)-style classifier of Asaoka and Kudo, we make three contributions. First, we give a rigorous equivalence result showing that, for computational-basis label states, the original SWAP-test-based training objective can be estimated without ancilla and reference registers by directly measuring the readout subsystem. This simplification reduces circuit resources while preserving the same training signal. Second, we propose a lightweight hybrid extension: instead of using only the probability of the correct readout bitstring, we estimate the full readout distribution and apply a learnable temperature-and-bias post-processing followed by a softmax normalization. Across MNIST, KMNIST, and Fashion-MNIST (restricted to eight classes), this classical post-processing yields substantial and statistically significant improvements in evaluation accuracy, consistently across cross-entropy and mean-squared-error objectives. Finally, to interpret these gains we analyze Fisher-geometry summaries based on effective dimension and complementary Fisher-spectrum diagnostics for both the quantum models and the classical baselines. Empirically, the hybrid quantum model is associated with a more spectrally concentrated and anisotropic Fisher geometry, with Fisher mass concentrated in fewer dominant directions and with a substantially larger local Fisher scale after training.

•Apr 14, 2026•0 citations

Quantum-Enhanced Temporal Embeddings via a Hybrid Seq2Seq Architecture

Tien-Ching Hsieh, Yun‐Cheng Tsai, Samuel Yen-Chi Chen

This work investigates how shallow NISQ-compatible quantum layers can improve temporal representation learning in real-world sequential data. We develop a QLSTM Seq2Seq autoencoder in which a depth-1 variational quantum circuit is embedded inside each recurrent gate, shaping the geometry of the learned latent manifold. Evaluated on fourteen rolling S&P500 windows from 2022–2025, the quantum-enhanced encoder produces smoother trajectories, clearer regime transitions, and more stable sector-coherent clusters than a classical LSTM baseline. These geometric properties support the use of a Radial Basis Function (RBF) kernel for downstream portfolio allocation, where both RBF-Graph and RBF-DivMom strategies consistently outperform their classical counterparts in risk-adjusted terms. Analysis across periods shows that compressed manifolds favour concentrated allocation, while dispersed manifolds favour diversification, demonstrating that latent geometry serves as a regime indicator. The results highlight a practical role for shallow hybrid quantum-classical layers in NISQ-era sequence modelling, offering a reproducible pathway for improving temporal embeddings in finance and other data-limited, noise-sensitive domains.

arXiv:2602.11578•International Conference on Quantum Communications, Networking, and Computing (QCNC)•Apr 6, 2026•1 citation
Few-shot domain adaptation remains fragile under compound distribution shifts because sparse target supervision amplifies representation noise and destabilizes metric structure. This study introduces self-supervised contrastive learning directly in a quantum latent space produced by a variational quantum circuit encoder, followed by prototype-driven adaptation with distribution alignment. Across three target shifts (style, temporal sparsity, and mixed covariate plus conditional drift), the proposed method achieved mean episodic accuracies of 0.721, 0.705, and 0.689 at 1-shot, increasing to 0.832, 0.813, and 0.806 at 5-shot, and 0.851, 0.836, and 0.826 at 10-shot, respectively. At K=5, average accuracy improved from 0.672 under ERM to 0.824 with the full quantum-contrastive model, with 95% confidence intervals tightening from ±0.018 to ±0.011. Representation diagnostics showed a reduction in intra-class dispersion from 0.412 to 0.301 and an increase in inter-class separation from 0.688 to 0.824, while seed variability decreased from 0.012 to 0.007. Circuit ablations identified an optimal depth range of 6–8 layers, where accuracy peaked at 0.824–0.829 before declining at higher depth with increased episodic variance. Robustness tests under finite-shot measurement revealed graceful degradation: under severe shift, accuracy rose from 0.709 at 64 shots per observable to 0.801 at 2048 shots, with failure episodes reduced from 4.2% to 2.9% when increasing from 256 to 1024 shots. These results indicate that quantum-measurement-bounded embeddings can stabilize contrastive geometry and improve few-shot transfer when target labels are scarce.
Fine-grained image classification under limited labeled data presents a representation challenge: pretrained deep networks provide rich but high-dimensional features, whereas small-data learning and practical quantum encoding require substantially more compact representations. Direct dimensionality reduction can make such features computationally tractable but may not preserve the class-discriminative geometry needed for effective kernel learning. This study, therefore, examines whether supervised latent restructuring can produce a compact, class-aware, qubit-compatible representation that supports classical and quantum-kernel learning. EfficientNet-B0 embeddings were reduced from 1280 to 64 dimensions using PCA and subsequently restructured by LDA into an 11-dimensional Supervised Latent Restructuring (SLR) representation. SLR increased the held-out Silhouette coefficient from −0.0054 in PCA64 to 0.0490; however, predictive effects were model dependent: Random Forest and XGBoost improved, Linear SVM changed little, and RBF-SVM favored PCA64. A dimension-matched PCA11 control showed no consistent predictive advantage of SLR over PCA11 across the evaluated classifiers. Under the original 12-class taxonomy, the fixed quantum kernel remained below matched classical baselines. Under a broader three-group taxonomy, quantum-kernel performance improved with increasing training support, but RBF-SVM remained the strongest. $\gamma $ -only Quantum Kernel Alignment did not consistently outperform the fixed kernel and incurred up to $420\times $ greater runtime than the fixed kernel. Target-control and kernel-structure analyses showed that increased alignment produced little additional class-discriminative structure. Overall, the study shows that representation geometry, label granularity, numerical scaling, and adaptive feature scaling jointly influence small-data quantum-kernel learning; no quantum advantage was demonstrated.
Reliable decision-making with quantum neural networks on NISQ hardware requires uncertainty estimates that remain meaningful under finite-shot measurement noise and distribution shift. This study introduces an end-to-end pipeline that combines Bayesian variational quantum circuits with post hoc probability calibration and conformal prediction to produce coverage-controlled prediction sets. Across shot budgets from 128 to 4096 shots, the Bayesian variational circuit improves accuracy from 0.828 to 0.839 at 128 shots and from 0.862 to 0.892 at 4096 shots, while reducing negative log-likelihood from 0.462 to 0.448 at 128 shots and from 0.392 to 0.357 at 4096 shots. Calibration preserves accuracy while further lowering negative log-likelihood to 0.432 at 128 shots and 0.350 at 4096 shots, and reduces expected calibration error from 0.058 to 0.027. Conformalization achieves empirical coverage of 0.952 at miscoverage 0.05, with average set size 1.39 and singleton rate 0.71, and maintains coverage between 0.944 and 0.948 under feature corruption and subpopulation shift, where deterministic accuracy degrades to 0.801 and 0.774, respectively, while the Bayesian model retains 0.842 and 0.816. Ablations identify an operational knee at circuit depth 8, and show diminishing returns beyond 8 posterior samples per prediction, supporting deployment-efficient uncertainty control for quantum learning.
Quantum machine learning (QML) has attracted growing interest with the rapid parallel advances in large-scale classical machine learning and quantum technologies. Similar to classical machine learning, QML models also face challenges arising from the scarcity of labeled data, particularly as their scale and complexity increase. Here, we introduce self-supervised pretraining of quantum representations that reduces reliance on labeled data by learning invariances from unlabeled examples. We implement this paradigm on a programmable trapped-ion quantum computer, encoding images as quantum states. In situ contrastive pretraining on hardware yields a representation that, when fine-tuned, classifies image families with higher mean test accuracy and lower run-to-run variability than models trained from random initialization. Performance improvement is especially significant in regimes with limited labeled training data. We show that the learned invariances generalize beyond the pretraining image samples. Unlike prior work, our pipeline derives similarity from measured quantum overlaps and executes all training and classification stages on hardware. These results establish a label-efficient route to quantum representation learning, with direct relevance to quantum-native datasets and a clear path to larger classical inputs.
Supplementary Data Description This supplementary dataset provides the complete experimental results underlying Figures 4–9 of the manuscript. Each CSV file corresponds to one figure and reports mean performance values (and standard deviations where applicable), computed over multiple independent runs with different random seeds. Fig4_Low_Data_Performance.csv Contains classification accuracy (mean ± standard deviation) for Linear, MLP, and VQL heads across varying labeled target data fractions (1%, 5%, 10%, 20%). This dataset supports the low-data performance analysis presented in Fig. 4. Fig5_Parameter_Matched.csv Reports accuracy statistics for parameter-matched adaptation heads, including Linear Probe, MLP Adapter, Bottleneck Adapter, and the proposed VQL Head. This file enables fair comparison of adaptation strategies under comparable trainable parameter budgets (Fig. 5). Fig6_Domain_Shift.csv Provides source-domain accuracy, target-domain accuracy, and the resulting accuracy drop for each model under domain shift conditions. These results quantify robustness to distributional changes, as shown in Fig. 6. Fig7_Ablation.csv Contains ablation results evaluating the effect of quantum circuit depth and number of qubits on classification accuracy. This dataset supports the architectural analysis in Fig. 7. Fig8_Noise_Aware.csv Includes classification accuracy values under increasing quantum noise probabilities for all adaptation heads, enabling noise robustness evaluation corresponding to Fig. 8. Fig9_Convergence.csv Reports training accuracy as a function of training epochs for Linear, MLP, and VQL heads, illustrating convergence behavior and optimization stability (Fig. 9). Reproducibility Note All values are reported as mean performance over multiple independent runs, ensuring statistical reliability and reproducibility of the experimental findings.
Abstract Variational quantum algorithms are among the most widely studied approaches for applying near-term quantum devices to learning tasks. In this work we investigate a supervised quantum-autoencoder-style classifier that encodes images by amplitude encoding and predicts labels from a designated readout sub-register. Building on the quantum-autoencoder (QAE)-style classifier of Asaoka and Kudo, we make three contributions. First, we give a rigorous equivalence result showing that, for computational-basis label states, the original SWAP-test-based training objective can be estimated without ancilla and reference registers by directly measuring the readout subsystem. This simplification reduces circuit resources while preserving the same training signal. Second, we propose a lightweight hybrid extension: instead of using only the probability of the correct readout bitstring, we estimate the full readout distribution and apply a learnable temperature-and-bias post-processing followed by a softmax normalization. Across MNIST, KMNIST, and Fashion-MNIST (restricted to eight classes), this classical post-processing yields substantial and statistically significant improvements in evaluation accuracy, consistently across cross-entropy and mean-squared-error objectives. Finally, to interpret these gains we analyze Fisher-geometry summaries based on effective dimension and complementary Fisher-spectrum diagnostics for both the quantum models and the classical baselines. Empirically, the hybrid quantum model is associated with a more spectrally concentrated and anisotropic Fisher geometry, with Fisher mass concentrated in fewer dominant directions and with a substantially larger local Fisher scale after training.
This work investigates how shallow NISQ-compatible quantum layers can improve temporal representation learning in real-world sequential data. We develop a QLSTM Seq2Seq autoencoder in which a depth-1 variational quantum circuit is embedded inside each recurrent gate, shaping the geometry of the learned latent manifold. Evaluated on fourteen rolling S&P500 windows from 2022–2025, the quantum-enhanced encoder produces smoother trajectories, clearer regime transitions, and more stable sector-coherent clusters than a classical LSTM baseline. These geometric properties support the use of a Radial Basis Function (RBF) kernel for downstream portfolio allocation, where both RBF-Graph and RBF-DivMom strategies consistently outperform their classical counterparts in risk-adjusted terms. Analysis across periods shows that compressed manifolds favour concentrated allocation, while dispersed manifolds favour diversification, demonstrating that latent geometry serves as a regime indicator. The results highlight a practical role for shallow hybrid quantum-classical layers in NISQ-era sequence modelling, offering a reproducible pathway for improving temporal embeddings in finance and other data-limited, noise-sensitive domains.