pith. sign in

arxiv: 2606.12182 · v1 · pith:5F2PW5POnew · submitted 2026-06-10 · 💻 cs.LG · math.DS· math.OC

How Low Can You Go? Active Learning for Sparse Model Discovery in the Ultra-Low-Data Limit

Ana Larra\~naga , Urban Fasel , Steven L. Brunton This is my paper

Pith reviewed 2026-06-27 10:28 UTC · model grok-4.3

classification 💻 cs.LG math.DSmath.OC
keywords active learningSINDysparse model discoveryepistemic uncertaintydynamical systemsODE discoveryPDE discoveryultra-low data
0
0 comments X

The pith

Active learning with ensemble uncertainty recovers governing equations using far fewer data samples than random sampling.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper develops an active learning method to discover the sparse governing equations of dynamical systems when only very little data is available. It builds on the SINDy framework by using an ensemble approach called E-SINDy to measure uncertainty and then select new data points that are expected to be most helpful for identifying the correct model. Through tests on the Lorenz system and two PDEs, the method is shown to succeed with substantially less data than random selection in various noise conditions. A reader would care because acquiring data is often the bottleneck in scientific modeling, so smarter sampling could enable discoveries that would otherwise be too expensive.

Core claim

The authors claim that an active learning loop driven by epistemic uncertainty estimates from E-SINDy ensembles allows accurate recovery of the true sparse dynamics for both ODEs and PDEs in the ultra-low-data regime, outperforming random sampling across tested systems and noise levels.

What carries the argument

The E-SINDy ensemble, which generates multiple SINDy models to quantify epistemic uncertainty and thereby directs the selection of subsequent measurement locations.

If this is right

  • The method identifies correct governing equations for the Lorenz system with fewer points at different noise levels.
  • It distinguishes informative regions near shocks in Burgers' equation from less useful areas.
  • It navigates complex spatial patterns in the Kuramoto-Sivashinsky equation to find good sampling locations.
  • Overall accuracy in model discovery improves significantly compared to uniform random sampling.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • This sampling strategy could be adapted to guide sensor placement in real-world experiments with limited measurement budgets.
  • If the uncertainty signal proves robust, similar active learning could apply to other equation discovery techniques.
  • Performance may depend on the diversity of the initial ensemble models, suggesting a need to study initialization effects.

Load-bearing premise

The uncertainty scores produced by the E-SINDy ensemble point to data points that advance recovery of the actual governing equations rather than just areas of current model disagreement.

What would settle it

A direct comparison on the Lorenz system where, after the same number of actively chosen samples, the identified equations match the true ones no better than those from randomly selected samples.

Figures

Figures reproduced from arXiv: 2606.12182 by Ana Larra\~naga, Steven L. Brunton, Urban Fasel.

Figure 1
Figure 1. Figure 1: Active learning loop with E-SINDy for ODE dynamics discovery. Using the Lorenz system as an example, candidate initial conditions are sampled around the attractor and used to initialize short trajectories. An ensemble of SINDy models is trained on the current dataset, and the candidate with the highest predictive uncertainty is iteratively added to the training set until convergence, defined as agreement a… view at source ↗
Figure 2
Figure 2. Figure 2: Detailed description of the active learning sampling strategy outlined in the intermediate steps of [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Sampling strategy and its effect on the condition number κ(Θ) of the library matrix for the Hopf normal form oscillator. Sampling restricted to the limit cycle (r ≈ 1, blue) yields a poorly conditioned library matrix (κ(Θ) = 5 × 1010), as the trajectories explore only a narrow region of the state space, failing to capture the nonlinear dynamics away from the attractor. Sampling across a broader range of in… view at source ↗
Figure 4
Figure 4. Figure 4: Illustrative sampling strategies for the Kuramoto-Sivashinsky equation. Conventional approaches, including prior E-SINDy implementations, utilize the entire spatiotemporal domain, incorporating thousands of redundant points. Fixing a spatial location and sampling the full time series (orange, green, and red vertical lines) reduces the dataset size but retains substantial redundancy, as nearby spatial locat… view at source ↗
Figure 5
Figure 5. Figure 5: Comparison between active learning (blue) and random sampling (red) for dynamics discovery of the chaotic Lorenz system under 5% noise, evaluated via the ℓ0 and ℓ2 norms in the ultra-low data limit. The left panels show the median performance (dashed line) over 100 independent repetitions of the full loop, accounting for the variability introduced by the random selection of the initial candidate pool. The … view at source ↗
Figure 6
Figure 6. Figure 6: Evaluation of the noise effect on the active learning sampling strategy (blue) compared to random sampling (red) for the Lorenz system, across 1% (left), 5% (middle), and 10% (right) additive Gaussian noise levels. The first row shows the ℓ0 norm over the longer data budget range to show convergence, while the second row focuses on the ultra-low data limit (0 to 50 points). The third and fourth rows show t… view at source ↗
Figure 7
Figure 7. Figure 7: Location of initial conditions (ICs) in the training dataset for active learning (blue) versus random sampling (red) across different candidate pool sizes (c) for the Lorenz system. Solid and dashed contours enclose 50% and 90% of the ICs, respectively. Active learning consistently concentrates ICs in localized regions of the state space, in contrast to random sampling, which distributes them uniformly acr… view at source ↗
Figure 8
Figure 8. Figure 8: Analysis of sample placement for the two sampling strategies compared in this section: active learning via ensembling (blue) and D-optimal (grey). The left panels show a single representative run with the nominal data budget for Burgers (24 points, top) and KS (100 points, bottom). The right panels show the aggregated sample density over 100 independent repetitions, confirming the consistency of sample pla… view at source ↗
Figure 9
Figure 9. Figure 9: Evaluation of the ℓ2 norm (top) and relative residual error (bottom) for the three sampling strategies: ensemble-based active learning (blue), random sampling (red), and D-optimal (grey). The relative residual quantifies how well the identified PDE fits held￾out test data, computed as the mean absolute error between the library prediction Θξˆ and the measured time derivative ∂tu over a fixed held-out set T… view at source ↗
Figure 10
Figure 10. Figure 10: ℓ0 norm for Burgers and KS across the three sampling strategies: ensemble-based active learning (blue, left), random sampling (red, center), and D-optimal (grey, right). The top and bottom rows show results for the Burgers and Kuramoto–Sivashinsky equations, respectively. The data budget is shown up to the minimum number of points at which convergence was achieved. and uux rapidly, but the identification … view at source ↗
read the original abstract

Identifying the governing equations of complex dynamical systems remains a fundamental challenge across science and engineering. While early approaches relied on empirical data and heuristics, modern data-driven methods offer greater flexibility and fewer assumptions. However, data acquisition in real-world settings is often expensive. This work addresses this challenge by introducing an active learning strategy for dynamics discovery in the ultra-low data limit. Rather than sampling randomly, our method iteratively prioritizes regions that are most informative for model identification. This approach builds on Sparse Identification of Nonlinear Dynamics (SINDy), and utilizes an ensemble extension, E-SINDy, to estimate epistemic uncertainty and guide the sampling for both ordinary and partial differential equations (ODEs/PDEs). For ODEs, an exhaustive analysis is conducted on the Lorenz system across varying data budgets and noise levels. For PDEs, two systems with contrasting dynamical characteristics are examined: the Burgers' equation, where a sharp shock front creates a distinction between informative and uninformative regions, and the Kuramoto-Sivashinsky equation, which presents a more spatially complex sampling landscape. Across all scenarios, the proposed method accurately identifies the governing dynamics with significantly fewer data samples than random sampling.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper introduces an active learning strategy for sparse model discovery using an ensemble extension of SINDy (E-SINDy) to estimate epistemic uncertainty and prioritize informative data points in the ultra-low-data regime. It claims that this approach identifies governing dynamics for ODEs (Lorenz system) and PDEs (Burgers' and Kuramoto-Sivashinsky equations) with significantly fewer samples than random sampling, across varying data budgets and noise levels.

Significance. If the uncertainty-guided sampling is shown to preferentially advance recovery of the true sparse coefficients (rather than merely reducing ensemble variance), the method could meaningfully lower data requirements for equation discovery in settings where measurements are expensive. The extension of established SINDy frameworks with an explicit sampling loop is a natural direction, but its value hinges on empirical validation of the acquisition function's alignment with ground-truth term recovery.

major comments (2)
  1. [Abstract] Abstract: the central claim that the method 'accurately identifies the governing dynamics with significantly fewer data samples than random sampling' across all scenarios rests on the unverified assumption that E-SINDy epistemic uncertainty selects points whose addition enables sparse regression to recover the exact governing terms. No derivation, ablation, or quantitative comparison is provided showing that the acquisition function reduces coefficient error to the ground-truth model rather than simply shrinking spread among incorrect library models; high ensemble disagreement can arise from under-determined estimation or noise without supplying information about missing nonlinear terms.
  2. [Abstract] The manuscript provides no explicit sampling criteria, hyperparameter settings for the ensemble, or quantitative results tables that would allow verification of the exhaustive testing asserted in the abstract; without these, the support for superior performance over random sampling cannot be assessed and may involve unstated post-hoc choices.
minor comments (2)
  1. Notation for the acquisition function and how it is computed from the E-SINDy ensemble variance should be clarified with an explicit equation.
  2. The description of the Lorenz, Burgers', and Kuramoto-Sivashinsky test cases would benefit from a table summarizing the exact data budgets, noise levels, and performance metrics (e.g., term recovery rate or coefficient error) for active vs. random sampling.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which help clarify the presentation and validation of our active learning approach for sparse model discovery. We address each major comment below and commit to revisions that strengthen the empirical support for the method's claims.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that the method 'accurately identifies the governing dynamics with significantly fewer data samples than random sampling' across all scenarios rests on the unverified assumption that E-SINDy epistemic uncertainty selects points whose addition enables sparse regression to recover the exact governing terms. No derivation, ablation, or quantitative comparison is provided showing that the acquisition function reduces coefficient error to the ground-truth model rather than simply shrinking spread among incorrect library models; high ensemble disagreement can arise from under-determined estimation or noise without supplying information about missing nonlinear terms.

    Authors: We agree that distinguishing between uncertainty reduction and true term recovery is critical for validating the acquisition function. Our reported results on the Lorenz, Burgers', and Kuramoto-Sivashinsky systems show that the active sampling strategy recovers the exact ground-truth sparse coefficients (i.e., the correct library terms with accurate values) at lower data budgets than random sampling, across noise levels. This is quantified via successful model identification rates and coefficient error metrics relative to the known governing equations. However, to more explicitly demonstrate that the epistemic uncertainty from E-SINDy preferentially selects points that advance recovery of the true terms (rather than merely reducing ensemble variance among incorrect models), we will add a new ablation analysis. This will include plots and tables tracking coefficient error to ground truth versus ensemble disagreement as a function of selected samples, confirming alignment with term recovery. revision: yes

  2. Referee: [Abstract] The manuscript provides no explicit sampling criteria, hyperparameter settings for the ensemble, or quantitative results tables that would allow verification of the exhaustive testing asserted in the abstract; without these, the support for superior performance over random sampling cannot be assessed and may involve unstated post-hoc choices.

    Authors: The full manuscript details the sampling criteria (prioritizing points with highest epistemic uncertainty estimated via E-SINDy ensemble disagreement), ensemble hyperparameters (e.g., number of models, bootstrap variations, and library construction), and quantitative results (success rates, coefficient errors, and comparisons to random sampling) in the methods and results sections, with figures and tables for the Lorenz system across budgets/noise and the two PDEs. To address the concern about verifiability in support of the abstract claims, we will add an explicit summary table of all hyperparameters and sampling criteria, along with consolidated quantitative result tables, ensuring all experimental settings are fully specified without post-hoc adjustments. revision: yes

Circularity Check

0 steps flagged

No significant circularity; method extends prior frameworks with independent empirical validation

full rationale

The paper proposes an active learning loop that uses E-SINDy ensemble uncertainty to select new samples for SINDy regression. Performance claims rest on numerical experiments (Lorenz, Burgers, Kuramoto-Sivashinsky) comparing active vs. random sampling, not on any derivation that reduces by construction to fitted parameters or self-referential definitions. Prior SINDy/E-SINDy citations supply the base regressor but are not invoked as a uniqueness theorem or ansatz that forces the new result; the active sampling strategy is evaluated separately against ground-truth recovery metrics.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the standard SINDy assumption that a pre-chosen library of candidate functions is rich enough to contain the true terms, plus the untested premise that ensemble uncertainty correlates with informativeness for model identification. No new entities are introduced and no free parameters are explicitly fitted in the abstract.

free parameters (1)
  • SINDy sparsity threshold
    Standard hyperparameter in SINDy whose value is not specified in the abstract but is required for the sparse regression step.
axioms (1)
  • domain assumption The candidate function library contains the true terms of the governing equations.
    Implicit foundation of all SINDy-based methods; invoked when the method claims to recover the correct dynamics.

pith-pipeline@v0.9.1-grok · 5752 in / 1297 out tokens · 38094 ms · 2026-06-27T10:28:43.197645+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

62 extracted references · 37 canonical work pages

  1. [1]

    Brunton and J

    Steven L. Brunton and J. Nathan Kutz.Data-Driven Science and Engineering: Machine Learning, Dynami- cal Systems, and Control. Cambridge University Press, 2019

    2019

  2. [2]

    When is a system discoverable from data? discovery requires chaos, 2025

    Zakhar Shumaylov, Peter Zaika, Philipp Scholl, Gitta Kutyniok, Lior Horesh, and Carola-Bibiane Schönlieb. When is a system discoverable from data? discovery requires chaos, 2025. URLhttps: //arxiv.org/abs/2511.08860

    arXiv 2025

  3. [3]

    Brunton, Joshua L

    Steven L. Brunton, Joshua L. Proctor, and J. Nathan Kutz. Discovering governing equations from data by sparse identification of nonlinear dynamical systems.Proceedings of the National Academy of Sciences, 113(15):3932–3937, 2016. doi: 10.1073/pnas.1517384113

    work page doi:10.1073/pnas.1517384113 2016

  4. [4]

    Learning from noisy examples.Machine Learning, 2(4):343–370, 1988

    Dana Angluin and Philip Laird. Learning from noisy examples.Machine Learning, 2(4):343–370, 1988. ISSN 1573-0565. doi: 10.1007/BF00116829

    work page doi:10.1007/bf00116829 1988

  5. [5]

    Cheeseman, Ivo F

    Suryanarayana Maddu, Bevan L. Cheeseman, Ivo F. Sbalzarini, and Christian L. Müller. Stability selection enables robust learning of differential equations from limited noisy data.Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 478(2262):20210916, 06 2022. ISSN 1364-

    2022

  6. [6]

    doi: 10.1098/rspa.2021.0916

    work page doi:10.1098/rspa.2021.0916 2021

  7. [7]

    Nghiem, Thomas Beckers, Mahyar Fazlyab, Enrique Mallada, Colin Jones, Dra- guna Vrabie, Steven L

    Jan Drgo ˇ na, Truong X. Nghiem, Thomas Beckers, Mahyar Fazlyab, Enrique Mallada, Colin Jones, Dra- guna Vrabie, Steven L. Brunton, and Rolf Findeisen. Safe physics-informed machine learning for dy- namics and control, 2025. URLhttps://arxiv.org/abs/2504.12952

    arXiv 2025

  8. [8]

    Nathan Kutz, and Steven L

    Kathleen Champion, Bethany Lusch, J. Nathan Kutz, and Steven L. Brunton. Data-driven discovery of coordinates and governing equations.Proceedings of the National Academy of Sciences, 116(45):22445– 22451, 2019. doi: 10.1073/pnas.1906995116

    work page doi:10.1073/pnas.1906995116 2019

  9. [9]

    Andrew Wagenmaker.A Theory of Active Learning in Dynamic Environments. Ph.d. thesis, University of Washington, 2024. URLhttps://digital.lib.washington.edu/researchworks/items/ 8800ffc0-a15a-4817-aa05-2c173d1f098c/full. University of Washington ResearchWorks

    2024

  10. [10]

    Adaptive sampling methods for learning dynamical systems

    Zichen Zhao and Qianxiao Li. Adaptive sampling methods for learning dynamical systems. InPro- ceedings of Mathematical and Scientific Machine Learning, volume 190 ofProceedings of Machine Learning Research, pages 335–350. PMLR, 2022

    2022

  11. [11]

    Small data machine learning in materials science

    Pengcheng Xu, Xiaobo Ji, Minjie Li, and Wencong Lu. Small data machine learning in materials science. npj Computational Materials, 9(1):42, 2023. ISSN 2057-3960. doi: 10.1038/s41524-023-01000-z

    work page doi:10.1038/s41524-023-01000-z 2023

  12. [12]

    Ethan Pickering, Stephen Guth, George Em Karniadakis, and Themistoklis P . Sapsis. Discovering and forecasting extreme events via active learning in neural operators.Nature Computational Science, 2(12): 823–833, 2022. doi: 10.1038/s43588-022-00376-0

    work page doi:10.1038/s43588-022-00376-0 2022

  13. [13]

    Hu, Enrico Camporeale, and B

    A. Hu, Enrico Camporeale, and B. Swiger. Multi-hour-ahead Dst index prediction using multi-fidelity boosted neural networks.Space Weather, 21(4):e2022SW003286, 2023. doi: 10.1029/2022SW003286. 16

    work page doi:10.1029/2022sw003286 2023

  14. [14]

    Integrating scientific knowledge with machine learning for engineering and environmental systems.ACM Comput

    Jared Willard, Xiaowei Jia, Shaoming Xu, Michael Steinbach, and Vipin Kumar. Integrating scientific knowledge with machine learning for engineering and environmental systems.ACM Comput. Surv., 55(4), 2022. ISSN 0360-0300. doi: 10.1145/3514228

    work page doi:10.1145/3514228 2022

  15. [15]

    Learning latent representations in high-dimensional state spaces using polynomial manifold constructions, 2023

    Rudy Geelen, Laura Balzano, and Karen Willcox. Learning latent representations in high-dimensional state spaces using polynomial manifold constructions, 2023. URLhttps://arxiv.org/abs/2306. 13748

    2023

  16. [16]

    Patsatzis, Lucia Russo, and Constantinos Siettos

    Alessandro Della Pia, Dimitrios G. Patsatzis, Lucia Russo, and Constantinos Siettos. Learning the latent dynamics of fluid flows from high-fidelity numerical simulations using parsimonious diffusion maps.Physics of Fluids, 36(10):105187, 10 2024. ISSN 1070-6631. doi: 10.1063/5.0232378

    work page doi:10.1063/5.0232378 2024

  17. [17]

    Active learning for neural PDE solvers, 2024

    Daniel Musekamp, Marimuthu Kalimuthu, David Holzmüller, Makoto Takamoto, and Mathias Niepert. Active learning for neural PDE solvers, 2024. URLhttps://arxiv.org/abs/2408. 01536

    2024

  18. [18]

    Pool-based sequential active learning for regression.IEEE Transactions on Neural Networks and Learning Systems, 30(5):1348–1359, 2019

    Dongrui Wu. Pool-based sequential active learning for regression.IEEE Transactions on Neural Networks and Learning Systems, 30(5):1348–1359, 2019. doi: 10.1109/TNNLS.2018.2868649

    work page doi:10.1109/tnnls.2018.2868649 2019

  19. [19]

    Extracting structured dynamical sys- tems using sparse optimization with very few samples.Multiscale Modeling & Simulation, 18(4):1435– 1461, 2020

    Hayden Schaeffer, Giang Tran, Rachel Ward, and Linan Zhang. Extracting structured dynamical sys- tems using sparse optimization with very few samples.Multiscale Modeling & Simulation, 18(4):1435– 1461, 2020. doi: 10.1137/18M1194730

    work page doi:10.1137/18m1194730 2020

  20. [20]

    Query by committee

    Hyunjune Sebastian Seung, Manfred Opper, and Haim Sompolinsky. Query by committee. InPro- ceedings of the Fifth Annual Workshop on Computational Learning Theory, pages 287–294. ACM, 1992. doi: 10.1145/130385.130417

    work page doi:10.1145/130385.130417 1992

  21. [21]

    Nathan Kutz, Bingni W

    Urban Fasel, J. Nathan Kutz, Bingni W. Brunton, and Steven L. Brunton. Ensemble-SINDy: Robust sparse model discovery in the low-data, high-noise limit, with active learning and control.Proceedings of the Royal Society A, 478(2260):20210904, 2022. doi: 10.1098/rspa.2021.0904

    work page doi:10.1098/rspa.2021.0904 2022

  22. [22]

    Fisher.The Design of Experiments

    Ronald A. Fisher.The Design of Experiments. Oliver and Boyd, 1937. URLhttps://books.google. com/books?id=IMrtAAAAMAAJ

    1937

  23. [23]

    Loudon.The Quantum Theory of Light

    Anthony Curtis Atkinson and Alexander Nikolaev Donev.Optimum Experimental Designs. Oxford University Press, 08 1992. ISBN 9780198522546. doi: 10.1093/oso/9780198522546.001.0001

    work page doi:10.1093/oso/9780198522546.001.0001 1992

  24. [24]

    Optimum designs in regression problems.The annals of mathematical statistics, 30(2):271–294, 1959

    Jack Kiefer and Jacob Wolfowitz. Optimum designs in regression problems.The annals of mathematical statistics, 30(2):271–294, 1959

    1959

  25. [25]

    2006.Decision Modelling For Health Economic Evaluation

    Anthony C. Atkinson, Alexander Nikolaev Donev, and Randall D. Tobias.Optimum Experimental Designs, with SAS. Oxford University Press, 05 2007. ISBN 9780199296590. doi: 10.1093/oso/ 9780199296590.001.0001

    work page doi:10.1093/oso/ 2007

  26. [26]

    George E. P . Box.Sequential Experimentation and Sequential Assembly of Designs. Report (University of Wisconsin–Madison. Center for Quality and Productivity Improvement). Center for Quality and Productivity Improvement, University of Wisconsin-Madison, 1992. URLhttps://books.google. com/books?id=UWnDNAAACAAJ

    1992

  27. [27]

    Automl based workflow for design of ex- periments (doe) selection: benchmarking data acquisition strategies with simulation models.Scientific Reports, 14(1):32170, 2024

    Xukuan Xu, Donghui Li, Jinghou Bi, and Michael Moeckel. Automl based workflow for design of ex- periments (doe) selection: benchmarking data acquisition strategies with simulation models.Scientific Reports, 14(1):32170, 2024. doi: 10.1038/s41598-024-83581-3

    work page doi:10.1038/s41598-024-83581-3 2024

  28. [28]

    Synthesis Lectures on Artificial Intelligence and Machine Learning

    Burr Settles.Active Learning. Synthesis Lectures on Artificial Intelligence and Machine Learning. Springer, Cham, 2012. doi: 10.1007/978-3-031-01560-1. 17

    work page doi:10.1007/978-3-031-01560-1 2012

  29. [29]

    A framework and bench- mark for deep batch active learning for regression, 2023

    David Holzmüller, Viktor Zaverkin, Johannes Kästner, and Ingo Steinwart. A framework and bench- mark for deep batch active learning for regression, 2023. URLhttps://arxiv.org/abs/2203. 09410

    2023

  30. [30]

    A survey on instance selection for active learning.Knowledge and Information Systems, 35(2):249–283, 2013

    Yifan Fu, Xingquan Zhu, and Bin Li. A survey on instance selection for active learning.Knowledge and Information Systems, 35(2):249–283, 2013. doi: 10.1007/s10115-012-0507-8

    work page doi:10.1007/s10115-012-0507-8 2013

  31. [31]

    Cardenas, and Nick Dexter

    Ben Adcock, Juan M. Cardenas, and Nick Dexter. CS4ML: A general framework for active learning with arbitrary data based on christoffel functions, 2023. URLhttps://arxiv.org/abs/2306. 00945

    2023

  32. [32]

    Active learning for nonlinear system identifica- tion with guarantees.Journal of Machine Learning Research, 23(1), 2022

    Horia Mania, Michael I Jordan, and Benjamin Recht. Active learning for nonlinear system identifica- tion with guarantees.Journal of Machine Learning Research, 23(1), 2022

    2022

  33. [33]

    Shields, Kurtis Gurley, Ryan Catarelli, Mohit Chauhan, Mariel Ojeda-Tuz, and Forrest J

    Michael D. Shields, Kurtis Gurley, Ryan Catarelli, Mohit Chauhan, Mariel Ojeda-Tuz, and Forrest J. Masters. Active learning applied to automated physical systems increases the rate of discovery.Scien- tific Reports, 13(1):8402, 2023. doi: 10.1038/s41598-023-35257-7

    work page doi:10.1038/s41598-023-35257-7 2023

  34. [34]

    Deep active learning for nonlinear system identification, 2023

    Erlend Torje Berg Lundby, Adil Rasheed, Ivar Johan Halvorsen, Dirk Reinhardt, Sebastien Gros, and Jan Tommy Gravdahl. Deep active learning for nonlinear system identification, 2023. URLhttps: //arxiv.org/abs/2302.12667

    arXiv 2023

  35. [35]

    Hypergraph reconstruc- tion from dynamics.Nature Communications, 16(1):2691, 2025

    Robin Delabays, Giulia De Pasquale, Florian Dörfler, and Yuanzhao Zhang. Hypergraph reconstruc- tion from dynamics.Nature Communications, 16(1):2691, 2025. doi: 10.1038/s41467-025-57664-2. URL https://doi.org/10.1038/s41467-025-57664-2

    work page doi:10.1038/s41467-025-57664-2 2025

  36. [36]

    Jonathan Horrocks and Chris T. Bauch. Algorithmic discovery of dynamic models from infectious disease data.Scientific Reports, 10(1):7061, 2020. doi: 10.1038/s41598-020-63877-w. URLhttps: //doi.org/10.1038/s41598-020-63877-w

    work page doi:10.1038/s41598-020-63877-w 2020

  37. [37]

    Nathan Kutz, and Steven L

    Eurika Kaiser, J. Nathan Kutz, and Steven L. Brunton. Sparse identification of nonlinear dynamics for model predictive control in the low-data limit.Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 474(2219):20180335, 11 2018. ISSN 1364-5021. doi: 10.1098/rspa.2018.0335

    work page doi:10.1098/rspa.2018.0335 2018

  38. [38]

    Nathan Kutz, and Steven L

    Nicholas Zolman, Christian Lagemann, Urban Fasel, J. Nathan Kutz, and Steven L. Brunton. SINDy- RL for interpretable and efficient model-based reinforcement learning.Nature Communications, 16(1): 10714, 2025. doi: 10.1038/s41467-025-65738-4

    work page doi:10.1038/s41467-025-65738-4 2025

  39. [39]

    Active learning and bayesian optimization: A unified perspective to learn with a goal.Archives of Computational Methods in Engineering, 31(5): 2985–3013, 2024

    Francesco Di Fiore, Michela Nardelli, and Laura Mainini. Active learning and bayesian optimization: A unified perspective to learn with a goal.Archives of Computational Methods in Engineering, 31(5): 2985–3013, 2024. doi: 10.1007/s11831-024-10064-z

    work page doi:10.1007/s11831-024-10064-z 2024

  40. [40]

    Consi, Luca Bonfiglio, Yue Ma, Lena R

    Dixia Fan, Grégoire Jodin, Thomas R. Consi, Luca Bonfiglio, Yue Ma, Lena R. Keyes, George E. Kar- niadakis, and Michael S. Triantafyllou. A robotic intelligent towing tank for learning complex fluid- structure dynamics.Science Robotics, 4(36):eaay5063, 2019. doi: 10.1126/scirobotics.aay5063

    work page doi:10.1126/scirobotics.aay5063 2019

  41. [41]

    Neural network ensembles, cross validation, and active learning

    Anders Krogh and Jesper Vedelsby. Neural network ensembles, cross validation, and active learning. In G. Tesauro, D. Touretzky, and T. Leen, editors,Advances in Neural Information Processing Systems, volume 7. MIT Press, 1994

    1994

  42. [42]

    Springer-Verlag, 2005

    Vladimir Vovk, Alex Gammerman, and Glenn Shafer.Algorithmic Learning in a Random World. Springer-Verlag, 2005. ISBN 0387001522

    2005

  43. [43]

    Sparse identification of nonlinear dynamics with conformal prediction, 2025

    Urban Fasel. Sparse identification of nonlinear dynamics with conformal prediction, 2025. URL https://arxiv.org/abs/2507.11739. 18

    arXiv 2025

  44. [44]

    Lloyd Fung, Urban Fasel, and Matthew Juniper. Rapid bayesian identification of sparse nonlinear dynamics from scarce and noisy data.Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 481(2307):20240200, 2025. ISSN 1364-5021. doi: 10.1098/rspa.2024.0200

    work page doi:10.1098/rspa.2024.0200 2025

  45. [45]

    Rudy, Steven L

    Samuel H. Rudy, Steven L. Brunton, Joshua L. Proctor, and J. Nathan Kutz. Data-driven discovery of partial differential equations.Science Advances, 3(4):e1602614, 2017. doi: 10.1126/sciadv.1602614

    work page doi:10.1126/sciadv.1602614 2017

  46. [46]

    Noack, and Steven L

    Jean-Christophe Loiseau, Bernd R. Noack, and Steven L. Brunton. Sparse reduced-order modelling: sensor-based dynamics to full-state estimation.Journal of Fluid Mechanics, 844:459–490, 2018. doi: 10.1017/jfm.2018.147

    work page doi:10.1017/jfm.2018.147 2018

  47. [47]

    Messenger and David M

    Daniel A. Messenger and David M. Bortz. Weak sindy: Galerkin-based data-driven model selection. Multiscale Modeling & Simulation, 19(3):1474–1497, 2021. doi: 10.1137/20M1343166

    work page doi:10.1137/20m1343166 2021

  48. [48]

    Patrick A. K. Reinbold, Daniel R. Gurevich, and Roman O. Grigoriev. Using noisy or incomplete data to discover models of spatiotemporal dynamics.Phys. Rev. E, 101:010203, 2020. doi: 10.1103/PhysRevE. 101.010203

    work page doi:10.1103/physreve 2020

  49. [49]

    Hayden Schaeffer and Scott G. McCalla. Sparse model selection via integral terms.Phys. Rev. E, 96: 023302, Aug 2017. doi: 10.1103/PhysRevE.96.023302

    work page doi:10.1103/physreve.96.023302 2017

  50. [50]

    Kaptanoglu, Jared L

    Alan A. Kaptanoglu, Jared L. Callaham, Aleksandr Aravkin, Christopher J. Hansen, and Steven L. Brunton. Promoting global stability in data-driven models of quadratic nonlinear dynamics.Phys. Rev. Fluids, 6:094401, 2021. doi: 10.1103/PhysRevFluids.6.094401

    work page doi:10.1103/physrevfluids.6.094401 2021

  51. [51]

    A robust sindy approach by combining neural net- works and an integral form, 2023

    Ali Forootani, Pawan Goyal, and Peter Benner. A robust sindy approach by combining neural net- works and an integral form, 2023. URLhttps://arxiv.org/abs/2309.07193

    arXiv 2023

  52. [52]

    Diverse ensembles for active learning

    Prem Melville and Raymond J Mooney. Diverse ensembles for active learning. InProceedings of the twenty-first international conference on Machine learning, page 74, 2004

    2004

  53. [53]

    An Entropy Stable High-Order Discontinuous Galerkin Method on Cut Meshes

    Wenhan Gao and Chunmei Wang. Active learning based sampling for high-dimensional nonlinear partial differential equations.Journal of Computational Physics, 475:111848, 2023. doi: 10.1016/j.jcp. 2022.111848

    work page doi:10.1016/j.jcp 2023

  54. [54]

    Mohamad and Themistoklis P

    Mustafa A. Mohamad and Themistoklis P . Sapsis. Sequential sampling strategy for extreme event statistics in nonlinear dynamical systems.Proceedings of the National Academy of Sciences, 115(44):11138– 11143, 2018. doi: 10.1073/pnas.1813263115

    work page doi:10.1073/pnas.1813263115 2018

  55. [55]

    On fast leverage score sampling and optimal learning

    Alessandro Rudi, Daniele Calandriello, Luigi Carratino, and Lorenzo Rosasco. On fast leverage score sampling and optimal learning. In S. Bengio, H. Wallach, H. Larochelle, K. Grauman, N. Cesa-Bianchi, and R. Garnett, editors,Advances in Neural Information Processing Systems, volume 31. Curran Asso- ciates, Inc., 2018

    2018

  56. [56]

    Society for Industrial and Applied Mathematics,

    Friedrich Pukelsheim.Optimal Design of Experiments. Society for Industrial and Applied Mathematics,

  57. [57]

    doi: 10.1137/1.9780898719109

    work page doi:10.1137/1.9780898719109

  58. [58]

    Manohar, B

    Krithika Manohar, Bingni W. Brunton, J. Nathan Kutz, and Steven L. Brunton. Data-driven sparse sensor placement for reconstruction: Demonstrating the benefits of exploiting known patterns.IEEE Control Systems Magazine, 38(3):63–86, 2018. doi: 10.1109/MCS.2018.2810460

    work page doi:10.1109/mcs.2018.2810460 2018

  59. [59]

    A mathematical model illustrating the theory of turbulence

    Johannes Martinus Burgers. A mathematical model illustrating the theory of turbulence. In Richard Von Mises and Theodore Von Kármán, editors,Advances in Applied Mechanics, volume 1, pages 171–199. Elsevier, 1948. doi: 10.1016/S0065-2156(08)70100-5

    work page doi:10.1016/s0065-2156(08)70100-5 1948

  60. [60]

    Sivashinsky

    Grigory I. Sivashinsky. Nonlinear analysis of hydrodynamic instability in laminar flames—I. Deriva- tion of basic equations.Acta Astronautica, 4:1177–1206, 1977. 19

    1977

  61. [61]

    Diffusion-induced chaos in reaction systems.Progress of Theoretical Physics Supple- ment, 64:346–367, 1978

    Yoshiki Kuramoto. Diffusion-induced chaos in reaction systems.Progress of Theoretical Physics Supple- ment, 64:346–367, 1978

    1978

  62. [62]

    Hyman and Basil Nicolaenko

    James M. Hyman and Basil Nicolaenko. The Kuramoto–Sivashinsky equation: A bridge between PDEs and dynamical systems.Physica D, 18:113–126, 1986. 20

    1986