quant-phFeb 16, 2026

Quantum Reservoir Computing with Neutral Atoms on a Small, Complex, Medical Dataset

Authors: Luke Antoncich, Yuben Moodley, Ugo Varetto, Jingbo Wang, Jonathan Wurtz, Jing Chen, Pascal Jahan Elahi, Casey R. Myers

Organizations: Centre for Quantum Information, Simulation and Algorithms, The University of Western Australia, Perth, Australia · Centre for Respiratory Health, University of Western Australia, Perth, WA, Australia · School of Medicine & Pharmacology, University of Western Australia, Perth, Western Australia, Australia · Dept of Respiratory Medicine, Fiona Stanley Hospital, Perth, WA, Australia · 

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Biomarker-based prediction of clinical outcomes is challenging due to nonlinear relationships, correlated features, and the limited size of many medical datasets. Classical machine-learning methods can struggle under these conditions, motivating the search for alternatives. In this work, we investigate quantum reservoir computing (QRC), using both noiseless emulation and hardware execution on the neutral-atom Rydberg processor Aquila. We evaluate performance with six classical machine-learning models and use SHAP to generate feature subsets. We find that models trained on emulated quantum features achieve mean test accuracies comparable to those trained on classical features, but have higher training accuracies and greater variability over data splits, consistent with overfitting. When comparing hardware execution of QRC to noiseless emulation, the models are more robust over different data splits and often exhibit statistically significant improvements in mean test accuracy. This combination of improved accuracy and increased stability is suggestive of a regularising effect induced by hardware execution. To investigate the origin of this behaviour, we examine the statistical differences between hardware and emulated quantum feature distributions. We find that hardware execution applies a structured, time-dependent transformation characterised by compression toward the mean and a progressive reduction in mutual information relative to emulation.

Figures & tables

Figure 1: Performance of six machine-learning models trained on SHAP-ranked classical feature subsets. We use five unique train/test splits of the data. For each model, data split, and feature count, we evaluate performance with optimal machine-learning hyperparameters. Violin plots show variability in test accuracy across the five data splits. The highest mean test accuracy is marked by the black star. The dotted line marks the mean test accuracy when using all 56 data features. In all cases the models exhibit improved performance when using a subset of all available data features.
Figure 2: Performance of six machine-learning models trained on quantum features generated by an emulated quantum reservoir. The same SHAP-ranked feature subsets as in the classical results are used to encode data into the atoms of the reservoir. We use five unique train/test splits of the data. For each model, data split, and atom count, we evaluate performance with optimal machine-learning hyperparameters. Violin plots show variability in test accuracy across the five data splits. The highest mean test accuracy is marked by the black star. The optimal performance achieved using emulated quantum features is comparable to that obtained with classical features. Models using emulated quantum features generally have higher training accuracies and larger variability across data splits, indicative of overfitting.
Figure 3: Changes in mean test accuracy (vertical axis) and split-to-split variability (horizontal axis) when moving from emulated to hardware quantum features. We observe that models trained on hardware quantum features often exhibit less variability across data splits than emulated quantum features. Importantly, this is often also accompanied by an increase in mean test accuracy. This indicates that executing the quantum reservoir on hardware helps to mitigate the overfitting tendencies observed with the emulated reservoir.
Figure 4: Standardised difference (see Eq. 2 ) in mean test accuracy between models trained on hardware and emulated quantum features. Points with a black outline correspond to models which also exhibited reduced variability on hardware compared to emulation. Across most model and feature-count combinations, hardware performance is comparable to or exceeds that of emulation, with several configurations exhibiting statistically significant advantages for hardware execution (standardised difference >2>2 ).
Figure 5: Comparison of hardware and emulated features across evolution times. Top left: Heatmap of residuals ( HW−Emu\mathrm{HW}-\mathrm{Emu} ), showing structured sample- and feature-dependent deviations. Top right: Mutual information between hardware and emulated features, decreasing at later times. Bottom row: Residuals plotted against emulated values, with ⟨Z⟩\langle Z\rangle and ⟨ZZ⟩\langle ZZ\rangle features shown separately. Overlaid points indicate equal-population bins (0.01–0.99 quantile range), showing bin means and variances. Residuals follow an approximately linear trend with Gaussian-distributed fluctuations within each bin.
Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

ModelHyperparameter Distributions
XGBoostn_estimators : randint(50, 200) learning_rate : uniform(0.01, 0.3) max_depth : randint(3, 10) min_child_weight : randint(1, 10) subsample : uniform(0.6, 0.4) colsample_bytree : uniform(0.6, 0.4) reg_alpha : uniform(0, 1) reg_lambda : uniform(1, 3)
Random Forestn_estimators : randint(50, 200) max_depth : {3,5,7,10,15,None}\{3,5,7,10,15,\text{None}\} min_samples_split : randint(2, 20) min_samples_leaf : randint(1, 10) max_features : {sqrt, log2, None}\{\text{sqrt},\,\text{log2},\,\text{None}\} bootstrap : {True, False}\{\text{True},\,\text{False}\}
Gradient Boostingn_estimators : randint(50, 200) learning_rate : uniform(0.01, 0.3) max_depth : randint(3, 8) min_samples_split : randint(2, 20) min_samples_leaf : randint(1, 10) subsample : uniform(0.6, 0.4)
Logistic RegressionC : uniform(0.01, 100) penalty : {l1, l2}\{\text{l1},\,\text{l2}\} solver : {liblinear, saga}\{\text{liblinear},\,\text{saga}\}
Support Vector Machine (SVM)C : uniform(0.1, 100) kernel : {linear, rbf}\{\text{linear},\,\text{rbf}\} gamma : {scale, auto, float}\{\text{scale},\,\text{auto},\,\text{float}\}
k-Nearest Neighbours (kNN)n_neighbors : randint(3, 20) weights : {uniform, distance}\{\text{uniform},\,\text{distance}\} algorithm : {auto, ball_tree, kd_tree, brute}\{\text{auto},\,\text{ball\_tree},\,\text{kd\_tree},\,\text{brute}\} p : {1,2}\{1,2\}
Appendix
Table 1: Hyperparameter distributions used for random search. randint(a,b) samples integers in [a,b][a,b] and uniform(a,b) samples real values in [a,a+b][a,a+b] . 1000 combinations were sampled at all stages of hyperparameter optimisation.
Figure 6: Mean SHAP ranking of the ten most important clinical features for each machine-learning model when trained using all data features. Each barplot reflects ranking averaged across all outer data splits. The six models naturally separate into two groups: the three tree-based models show strong agreement on one core subset of high-importance biomarkers, while Logistic Regression, SVM, and kNN agree on a different core subset. Feature 2 appears in all six models.
Figure 7: Model performance using hardware and emulated quantum features generated with aggregated SHAP feature encoding. Mean train and test accuracy are shown across feature counts, with violin plots indicating variability over data splits. Because the feature ranking is computed using information aggregated across all splits, these results are not directly comparable to the classical or emulated performance reported in the main text and should not be interpreted as unbiased performance estimates. Instead, the figure enables a controlled comparison between hardware and emulation under identical feature encodings.
More in this topic

Quantum Reservoir Computing

Quantum reservoir computing uses quantum dynamics to transform time-dependent inputs into features for prediction and classification.

See all 18 papers

Explore similar work

CardsList

  1. Quantum Reservoir Computing: Recent Advances and Future Directions

    Jul 20, 2026Shehbaz Tariq, Muhammad Talha, Arshid Ali +2Quantum Machine LearningReservoir Computing

    Interdisciplinary Centre for Security, Reliability and Trust (SnT), University of Luxembourg, Luxembourg City, Luxembourg · School of Computing, Engineering and Digital Technologies, Teesside University, Middlesbrough, United Kingdom

  2. Kernel-based optimization of measurement operators for quantum reservoir computers

    Feb 16, 2026Markus Gross, Hans-Martin RieserQuantum Reservoir ComputingQuantum Kernel Methods

    Institute for AI Safety and Security, German Aerospace Center (DLR), Sankt Augustin and Ulm, Germany

  3. Recurrent Quantum Feature Maps for Reservoir Computing

    Apr 3, 2026Utkarsh Singh, Aaron Z. Goldberg, Christoph Simon +1Quantum Reservoir ComputingQuantum Feature Maps

    Department of Physics, University of Ottawa, 25 Templeton Street, Ottawa, Ontario, K1N 6N5 Canada · National Research Council of Canada, 100 Sussex Drive, Ottawa, Ontario K1N 5A2, Canada · Institute for Quantum Science and Technology, Department of Physics and Astronomy, University of Calgary, Alberta T2N 1N4, Canada +1