Pith. sign in

REVIEW 3 major objections 2 minor 90 references

Linear oblique projections that balance sensitivity and state reconstruction enable data-driven forecasts and control of extreme events in chaotic systems.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.3

2026-06-27 23:08 UTC pith:4BNGHITI

load-bearing objection The paper swaps adjoints for backprop in CoBRAS, adds a local variant, and tests the combination on three chaotic systems for extreme-event forecasts and control. the 3 major comments →

arxiv 2606.05618 v1 pith:4BNGHITI submitted 2026-06-04 nlin.CD cs.LGmath.DS

Uncovering Extreme Event Mechanisms for Prediction and Control with Sensitivity-Balanced Projections

Nicholas Zolman , Sajeda Mokbel , Samuel E. Otto , Steven L. Brunton This is my paper
classification nlin.CD cs.LGmath.DS
keywords extreme eventschaotic dynamicssensitivity analysisdata-driven modelingoblique projectionsprediction and controlCoBRAS
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper develops the CoBRAS method to identify linear oblique projections that best capture the sensitivity of a quantity of interest while still allowing reconstruction of the original state. These projections are obtained via backpropagation rather than explicit adjoint calculations, and a local variant is introduced to handle spatially confined events. The approach is shown to produce simple forecast models that accurately predict extremes and to yield intuitive control laws that suppress them, with demonstrations across turbulent energy bursts, oscillator synchronization, and rogue wave formation. The method further extends to experimental or non-differentiable systems by training neural network surrogate models of the dynamics from data.

Core claim

The CoBRAS method identifies linear oblique projections that best capture the sensitivity of a quantity of interest and reconstruct the original state, enabling accurate data-driven forecasts and intuitive event suppression controllers across diverse chaotic systems. A local variant of CoBRAS produces spatially localized sensitivity-balanced projections. Backpropagation through automatically differentiable frameworks replaces cumbersome adjoint calculations, and neural network surrogates allow the same workflow on experimental data or systems not written in differentiable languages.

What carries the argument

The covariance balancing reduction using adjoint snapshots (CoBRAS) method, which computes linear oblique projections that jointly optimize sensitivity capture for a quantity of interest and faithful state reconstruction.

Load-bearing premise

Linear oblique projections obtained via backpropagation are sufficient to reveal the underlying instability mechanisms and support reliable prediction and control.

What would settle it

A demonstration that forecasts built from the CoBRAS projections perform no better than standard linear models on held-out data, or that the derived control laws fail to reduce extreme-event frequency in any of the three example systems, would falsify the central claim.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Simple linear models on the projections accurately predict extreme events in 2D Kolmogorov flow, FitzHugh-Nagumo networks, and modified nonlinear Schrödinger dynamics.
  • The sensitivity information in the projections directly yields control laws that suppress the targeted extreme events.
  • The local CoBRAS variant extends the same workflow to spatially localized phenomena without requiring global adjoint fields.
  • Neural-network surrogate models of the dynamics allow the entire pipeline to be applied to experimental data or black-box simulators.

Where Pith is reading between the lines

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

  • The same projection technique could be applied to high-dimensional climate or turbulence models where adjoint computation is prohibitive.
  • Replacing the linear forecast models with mildly nonlinear ones learned on the projected coordinates might further improve prediction horizons.
  • Because the method works from data alone, it offers a route to mechanism discovery in systems where the governing equations are only partially known.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 2 minor

Summary. The paper develops CoBRAS (covariance balancing reduction using adjoint snapshots) with backpropagation replacing adjoints to compute linear oblique projections that capture sensitivity of a quantity of interest while reconstructing the state. These projections are used to reveal mechanisms of extreme events, build data-driven forecasts, and design suppression controllers. A local variant is introduced for spatially localized events. Demonstrations are given on turbulent dissipation bursts in 2D Kolmogorov flow, spontaneous synchronization in FitzHugh-Nagumo oscillator networks, and rogue-wave formation in a modified nonlinear Schrödinger equation; the approach is further extended via neural-network surrogates to non-differentiable or experimental systems.

Significance. If the backpropagation-based sensitivities prove numerically stable and the resulting projections are shown to be predictive rather than fitted by construction, the work would offer a practical, interpretable route to mechanism discovery and control in high-dimensional chaotic systems where classical adjoints are cumbersome. The surrogate-model extension broadens applicability to laboratory data.

major comments (3)
  1. [§3] §3 (CoBRAS via backpropagation): the replacement of adjoint snapshots by automatic differentiation is presented without any discussion of gradient clipping, checkpointing, or horizon restriction. In the 2D Kolmogorov and modified NLS examples, positive Lyapunov exponents imply that sensitivities integrated over the times needed to observe extreme events will be dominated by exponential growth or decay; this directly threatens the claim that the resulting projections accurately capture the underlying instability mechanisms.
  2. [Results sections] Results sections for all three systems: the abstract states that the forecast models 'accurately predict extreme events' and that the mechanisms 'may be used to design control laws,' yet no quantitative metrics (prediction skill scores, ROC curves, false-alarm rates, or comparison against baseline linear or nonlinear predictors) are referenced. Without these, the central claim that the projections enable reliable prediction and control cannot be evaluated.
  3. [Local CoBRAS variant] Local CoBRAS variant: the modification for spatially localized events is introduced but the precise change to the covariance or sensitivity balancing step (e.g., weighting, masking, or localized inner product) is not specified, making it impossible to assess whether the variant preserves the original balancing property or merely approximates it.
minor comments (2)
  1. [Abstract and §3] The abstract and method description repeatedly use 'sensitivity-balanced projections' without an explicit equation defining the balancing objective (e.g., the precise form of the covariance or Gramian being balanced).
  2. [Figures] Figure captions for the three example systems should include the integration time horizon, number of trajectories, and any regularization applied during backpropagation so that reproducibility is immediate.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We are grateful to the referee for their constructive comments, which have prompted significant improvements to the manuscript. Below we provide point-by-point responses to the major comments, indicating the revisions made.

read point-by-point responses

  1. Referee: [§3] §3 (CoBRAS via backpropagation): the replacement of adjoint snapshots by automatic differentiation is presented without any discussion of gradient clipping, checkpointing, or horizon restriction. In the 2D Kolmogorov and modified NLS examples, positive Lyapunov exponents imply that sensitivities integrated over the times needed to observe extreme events will be dominated by exponential growth or decay; this directly threatens the claim that the resulting projections accurately capture the underlying instability mechanisms.

    Authors: We thank the referee for highlighting this important numerical consideration. The original submission indeed omitted explicit discussion of these implementation details and the potential impact of chaotic sensitivity growth. In the revised manuscript, we have expanded §3 to include a dedicated paragraph on numerical stability: we employ gradient clipping with a threshold of 1.0 during backpropagation, use checkpointing for longer horizons, and restrict the integration time to the minimal window required to observe the extreme event (typically 5-10 Lyapunov times in our examples). We further demonstrate through additional figures that the resulting sensitivity vectors remain aligned with the instability mechanisms rather than being overwhelmed by transient growth, by comparing against shorter-horizon computations. This revision directly addresses the concern and supports the validity of the projections. revision: yes

  2. Referee: [Results sections] Results sections for all three systems: the abstract states that the forecast models 'accurately predict extreme events' and that the mechanisms 'may be used to design control laws,' yet no quantitative metrics (prediction skill scores, ROC curves, false-alarm rates, or comparison against baseline linear or nonlinear predictors) are referenced. Without these, the central claim that the projections enable reliable prediction and control cannot be evaluated.

    Authors: The referee is correct that quantitative performance metrics were not provided in the original manuscript, which weakens the ability to evaluate the claims. We have revised the results sections to include: (i) ROC curves and area-under-curve values for extreme event prediction in each system, (ii) comparison of forecast skill against linear autoregressive models and simple threshold-based predictors, and (iii) false-alarm rates and precision-recall metrics. These additions confirm that the CoBRAS-based forecasts outperform the baselines, thereby substantiating the abstract claims. revision: yes

  3. Referee: [Local CoBRAS variant] Local CoBRAS variant: the modification for spatially localized events is introduced but the precise change to the covariance or sensitivity balancing step (e.g., weighting, masking, or localized inner product) is not specified, making it impossible to assess whether the variant preserves the original balancing property or merely approximates it.

    Authors: We agree that the description of the local variant was insufficiently precise. In the revised manuscript, we have clarified that the local CoBRAS is obtained by introducing a diagonal weighting matrix W (derived from a spatial mask around the event location) into the covariance and sensitivity Gramian computations: the balancing is performed on the weighted matrices C_w = W^{1/2} C W^{1/2} and similarly for the sensitivity, ensuring that the oblique projection remains optimal with respect to the localized inner product. This preserves the core balancing property while localizing the reduction. A mathematical derivation and pseudocode have been added to §3. revision: yes

Circularity Check

0 steps flagged

No significant circularity; method and demonstrations are self-contained

full rationale

The paper presents CoBRAS as an existing technique extended via backpropagation for sensitivity projections, then applies the resulting projections to build separate forecast models and controllers on three example systems. No equations or claims reduce a prediction to a fitted input by construction, and no load-bearing step relies on a self-citation chain that would make the central result tautological. The derivation chain (projection identification → reduced forecasts → control) remains independent of the input data fits.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides insufficient detail to enumerate free parameters, axioms, or invented entities; no explicit fitting, background assumptions, or new postulated objects are named.

pith-pipeline@v0.9.1-grok · 5812 in / 1086 out tokens · 17246 ms · 2026-06-27T23:08:13.944076+00:00 · methodology

0 comments
read the original abstract

Extreme events -- such as earthquakes and coronal mass ejections -- are common in many chaotic dynamical systems, yet are difficult to characterize and predict due to the subtle instability mechanisms that drive them. In this work, we develop an interpretable technique that reveals the underlying mechanisms behind extreme events and uses them to build data-driven forecasts and intuitive event suppression controllers. In particular, we utilize the covariance balancing reduction using adjoint snapshots (CoBRAS) method to identify linear oblique projections that best capture the sensitivity of a quantity of interest and reconstruct the original state. Importantly, we bypass the need for cumbersome adjoint calculations, instead using backpropagation via modern automatically differentiable numerical frameworks. To accommodate spatially localized events, we also introduce a new variant of CoBRAS to obtain local sensitivity-balanced projections. We demonstrate the utility of this approach to characterize extreme events across a diverse set of challenging systems, including turbulent bursts of energy dissipation in the 2D Kolmogorov Flow, spontaneous synchronization in networks of coupled FitzHugh-Nagumo oscillators, and the localized formation of ocean rogue waves from a modified nonlinear Schr\"odinger equation. For each example, we show that our simple forecast models accurately predict extreme events and that the underlying mechanisms may be used to design control laws to prevent these events. Finally, we demonstrate that by learning a neural network surrogate model of the dynamics directly from data, we may extend this approach to experimental systems and systems that are not natively written in an automatically differentiable programming language.

Figures

Figures reproduced from arXiv: 2606.05618 by Nicholas Zolman, Sajeda Mokbel, Samuel E. Otto, Steven L. Brunton.

Figure 1
Figure 1. Figure 1: Schematic of approach. (a) Obtaining the CoBRAS modes using forward snapshots and gradient samples from the Kolmogorov flow. (b) Projection onto the dominant Ψ modes of the Kolmogorov flow to reveal sensitive structure. (c) Examples of using the coordinates to identify mechanisms in FitzHugh-Nagumo oscillator networks, predicting event formation of rogue waves in the modified nonlinear Schrodinger equation… view at source ↗
Figure 2
Figure 2. Figure 2: Kolmogorov Flow (a) Energy dissipation, ε(t), of the Kolmogorov flow from t ∈ [0, 650] and vorticity snapshots taken from the shaded region during an extreme event t ∈ [515, 550]. (b) The first four POD and CoBRAS (Φ, Ψ) modes. (c) The projection onto the first two POD and CoBRAS modes; points are colored by whether the energy dissipation is above the energy threshold. (d) Comparison between the uncontroll… view at source ↗
Figure 3
Figure 3. Figure 3: FitzHugh-Nagumo Oscillators (a) The QoI for the FHN system over time exhibits small bursting with intermittent large extreme events. Dotted line corresponds to event threshold, q∗. (b) The first three POD and Ψ modes separated into the corresponding action on the x and y components. Dashed vertical line indicates node #23. (c) The projection onto the leading two POD and CoBRAS modes with states in orange i… view at source ↗
Figure 4
Figure 4. Figure 4: Modified Nonlinear Schr¨odinger Equation Overview (a) [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Modified Nonlinear Schr¨odinger Equation Results (a) [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Non-intrusive CoBRAS (a) Projection onto the first two dominant Co￾BRAS Ψ modes by obtaining gradients using autodiff through a learned FNO surrogate model. (b) The first four CoBRAS modes obtained from the FNO. (c) Comparison of the first two projected coordinates, z = ΨT ω, between CoBRAS modes obtained from a differentiable simulator vs the learned FNO surrogate. possible with adjoint-based or different… view at source ↗

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

90 extracted references · 3 canonical work pages

  1. [1]

    Epilepsy: Extreme events in the human brain

    Klaus Lehnertz. Epilepsy: Extreme events in the human brain. InExtreme events in nature and society, pages 123–

  2. [2]

    A bioelectrical phase tran- sition patterns the first vertebrate heartbeats.Nature, 622(7981):149–155, 2023

    Bill Z Jia, Yitong Qi, J David Wong-Campos, Sean G Mega- son, and Adam E Cohen. A bioelectrical phase tran- sition patterns the first vertebrate heartbeats.Nature, 622(7981):149–155, 2023

    2023

  3. [3]

    Elezgaray and A

    J. Elezgaray and A. Arneodo. Crisis-induced intermittent bursting in reaction-diffusion chemical systems.Phys. Rev. Lett, 68(5):714, 1992

    1992

  4. [4]

    Extreme events in gene regulatory networks with time-delays.Scientific Re- ports, 15(1):13064, 2025

    S Vinoth, S Leo Kingston, Sabarathinam Srinivasan, Suresh Kumarasamy, and Tomasz Kapitaniak. Extreme events in gene regulatory networks with time-delays.Scientific Re- ports, 15(1):13064, 2025

    2025

  5. [5]

    Extreme vortex-gust airfoil interactions at reynolds number 5000

    Kai Fukami, Luke Smith, and Kunihiko Taira. Extreme vortex-gust airfoil interactions at reynolds number 5000. Physical Review Fluids, 10(8):084703, 2025

    2025

  6. [6]

    Latif and N.S

    M. Latif and N.S. Keenlyside. El ni no/southern oscil- lation response to global warming.Proc. Natl. Acad. Sci, 106(49):20578–20583, 2009

    2009

  7. [7]

    Dijkstra.Nonlinear Climate Dynamics

    H.A. Dijkstra.Nonlinear Climate Dynamics. Cambridge University Press, Cambridge, UK, 2013

    2013

  8. [8]

    Roberts, J

    A. Roberts, J. Guckenheimer, E. Widiasih, A. Timmer- mann, and C.K.R.T. Jones. Mixed-mode oscillations of el ni no–southern oscillation.J. Atmos. Sci, 73(4):1755–1766, 2016

    2016

  9. [9]

    Extreme weather and climate events with ecological relevance: a review.Philosophical Transactions of the Royal Society B: Bio- logical Sciences, 372(1723):20160135, 2017

    Caroline C Ummenhofer and Gerald A Meehl. Extreme weather and climate events with ecological relevance: a review.Philosophical Transactions of the Royal Society B: Bio- logical Sciences, 372(1723):20160135, 2017

    2017

  10. [10]

    Extreme solar events.Living Reviews in Solar Physics, 19(1):2, 2022

    Edward W Cliver, Carolus J Schrijver, Kazunari Shibata, and Ilya G Usoskin. Extreme solar events.Living Reviews in Solar Physics, 19(1):2, 2022

    2022

  11. [11]

    Ex- treme events: Mechanisms and prediction.Applied Me- chanics Reviews, 71(5):050801, 2019

    Mohammad Farazmand and Themistoklis P Sapsis. Ex- treme events: Mechanisms and prediction.Applied Me- chanics Reviews, 71(5):050801, 2019

    2019

  12. [12]

    Statistics of extreme events in fluid flows and waves.Annual Review of Fluid Mechanics, 53(1):85–111, 2021

    Themistoklis P Sapsis. Statistics of extreme events in fluid flows and waves.Annual Review of Fluid Mechanics, 53(1):85–111, 2021

    2021

  13. [13]

    Springer, 2001

    Stuart Coles, Joanna Bawa, Lesley Trenner, and Pat Do- razio.An introduction to statistical modeling of extreme values, volume 208. Springer, 2001

    2001

  14. [14]

    Springer, 2009

    Amir Dembo.Large deviations techniques and applications. Springer, 2009

    2009

  15. [15]

    Computation of extreme values of time averaged observables in climate models with large deviation techniques.Journal of Statis- tical Physics, 179(5):1637–1665, 2020

    Francesco Ragone and Freddy Bouchet. Computation of extreme values of time averaged observables in climate models with large deviation techniques.Journal of Statis- tical Physics, 179(5):1637–1665, 2020

    2020

  16. [16]

    Predicting coronal mass ejections using machine learning methods.The As- trophysical Journal, 821(2):127, 2016

    Monica G Bobra and Stathis Ilonidis. Predicting coronal mass ejections using machine learning methods.The As- trophysical Journal, 821(2):127, 2016

    2016

  17. [17]

    Altwegg, V

    R. Altwegg, V . Visser, L.D. Bailey, and B. Erni. Learn- ing from single extreme events.Philos. Trans. R. Soc. B, 372(1723):20160141, 2017

    2017

  18. [18]

    Machine learn- ing predictors of extreme events occurring in complex dy- namical systems.Entropy, 21(10):925, 2019

    Stephen Guth and Themistoklis P Sapsis. Machine learn- ing predictors of extreme events occurring in complex dy- namical systems.Entropy, 21(10):925, 2019

    2019

  19. [19]

    Discovering and forecasting extreme events via active learning in neural operators.Na- ture Computational Science, 2(12):823–833, 2022

    Ethan Pickering, Stephen Guth, George Em Karniadakis, and Themistoklis P Sapsis. Discovering and forecasting extreme events via active learning in neural operators.Na- ture Computational Science, 2(12):823–833, 2022

    2022

  20. [20]

    Using machine learning to predict extreme events in complex systems.Proceedings of the National Academy of Sciences, 117(1):52–59, 2020

    Di Qi and Andrew J Majda. Using machine learning to predict extreme events in complex systems.Proceedings of the National Academy of Sciences, 117(1):52–59, 2020

    2020

  21. [21]

    Output- weighted and relative entropy loss functions for deep learning precursors of extreme events.Physica D: Nonlinear Phenomena, 443:133570, 2023

    Samuel H Rudy and Themistoklis P Sapsis. Output- weighted and relative entropy loss functions for deep learning precursors of extreme events.Physica D: Nonlinear Phenomena, 443:133570, 2023

    2023

  22. [22]

    Extreme event aware (η-) learning.arXiv preprint arXiv:2510.19161, 2025

    Kai Chang and Themistoklis P Sapsis. Extreme event aware (η-) learning.arXiv preprint arXiv:2510.19161, 2025

    work page arXiv 2025

  23. [23]

    Haller and S

    G. Haller and S. Wiggins. Multi-pulse jumping orbits and homoclinic trees in a modal truncation of the damped- forced nonlinear schr ¨odinger equation.Phys. D: Nonlinear Phenom, 85(3):311–347, 1995. 12

    1995

  24. [24]

    Haller and T

    G. Haller and T. Sapsis. Localized instability and attrac- tion along invariant manifolds.SIAM J. Appl. Dyn. Syst, 9(2):611–633, 2010

    2010

  25. [25]

    Farazmand and T.P

    M. Farazmand and T.P . Sapsis. Dynamical indicators for the prediction of bursting phenomena in high-dimensional systems.Phys. Rev. E, 94(3–1):032212, 2016

    2016

  26. [26]

    Farazmand

    M. Farazmand. An adjoint-based approach for finding in- variant solutions of navier–stokes equations.J. Fluid Mech, 795:278–312, 2016

    2016

  27. [27]

    Three-dimensional optimal perturbations in viscous shear flow.Physics of Flu- ids A: Fluid Dynamics, 4(8):1637–1650, 1992

    Kathryn M Butler and Brian F Farrell. Three-dimensional optimal perturbations in viscous shear flow.Physics of Flu- ids A: Fluid Dynamics, 4(8):1637–1650, 1992

    1992

  28. [28]

    Initial conditions for optimal growth in a coupled ocean–atmosphere model of enso.Journal of the atmospheric sciences, 55(4):537–557, 1998

    CJ Thompson. Initial conditions for optimal growth in a coupled ocean–atmosphere model of enso.Journal of the atmospheric sciences, 55(4):537–557, 1998

    1998

  29. [29]

    Farazmand and T.P

    M. Farazmand and T.P . Sapsis. A variational approach to probing extreme events in turbulent dynamical systems. Sci. Adv, 3(9):1701533, 2017

    2017

  30. [30]

    Ansmann, R

    G. Ansmann, R. Karnatak, K. Lehnertz, and U. Feudel. Ex- treme events in excitable systems and mechanisms of their generation.Phys. Rev. E, 88(5):052911, 2013

    2013

  31. [31]

    Massachusetts In- stitute of Technology, Department of Meteorology Cam- bridge, 1956

    Edward N Lorenz et al.Empirical orthogonal functions and statistical weather prediction, volume 1. Massachusetts In- stitute of Technology, Department of Meteorology Cam- bridge, 1956

    1956

  32. [32]

    The proper orthogonal decomposition in the analysis of turbu- lent flows.Ann

    Gal Berkooz, Philip Holmes, and John L Lumley. The proper orthogonal decomposition in the analysis of turbu- lent flows.Ann. Rev. Fluid Mech., 25(1):539–575, 1993

    1993

  33. [33]

    Grasping extreme aero- dynamics on a low-dimensional manifold.Nature Commu- nications, 14(1):6480, 2023

    Kai Fukami and Kunihiko Taira. Grasping extreme aero- dynamics on a low-dimensional manifold.Nature Commu- nications, 14(1):6480, 2023

    2023

  34. [34]

    Willcox and J

    K. Willcox and J. Peraire. Balanced model reduction via the proper orthogonal decomposition.AIAA Journal, 40(11):2323–2330, 2002

    2002

  35. [35]

    Model reduction for fluids, using bal- anced proper orthogonal decomposition.International Jour- nal of Bifurcation and Chaos, 15(03):997–1013, 2005

    Clarence W Rowley. Model reduction for fluids, using bal- anced proper orthogonal decomposition.International Jour- nal of Bifurcation and Chaos, 15(03):997–1013, 2005

    2005

  36. [36]

    Modeling of transi- tional channel flow using balanced proper orthogonal de- composition.Physics of Fluids, 20(3), 2008

    Milo ˇs Ilak and Clarence W Rowley. Modeling of transi- tional channel flow using balanced proper orthogonal de- composition.Physics of Fluids, 20(3), 2008

    2008

  37. [37]

    A critical-layer framework for turbulent pipe flow.Journal of Fluid Mechan- ics, 658:336–382, 2010

    Beverley J McKeon and Ati S Sharma. A critical-layer framework for turbulent pipe flow.Journal of Fluid Mechan- ics, 658:336–382, 2010

    2010

  38. [38]

    Op- position control within the resolvent analysis framework

    Mitul Luhar, Ati S Sharma, and Beverley J McKeon. Op- position control within the resolvent analysis framework. Journal of Fluid Mechanics, 749:597–626, 2014

    2014

  39. [39]

    Data-driven re- solvent analysis.Journal of Fluid Mechanics, 918:A10, 2021

    Benjamin Herrmann, Peter J Baddoo, Richard Semaan, Steven L Brunton, and Beverley J McKeon. Data-driven re- solvent analysis.Journal of Fluid Mechanics, 918:A10, 2021

    2021

  40. [40]

    Benner, P

    P . Benner, P . Goyal, and S. Gugercin.H 2-quasi-optimal model order reduction for quadratic-bilinear control sys- tems.SIAM J. Matrix Anal. & App., 39(2):983–1032, 2018

    2018

  41. [41]

    Balanced truncation for quadratic-bilinear control systems.Advances in Computa- tional Mathematics, 50(4):88, 2024

    Peter Benner and Pawan Goyal. Balanced truncation for quadratic-bilinear control systems.Advances in Computa- tional Mathematics, 50(4):88, 2024

    2024

  42. [42]

    Optimizing oblique projections for nonlinear systems using trajectories.SIAM Journal on Scientific Computing, 44(3):A1681–A1702, 2022

    Samuel E Otto, Alberto Padovan, and Clarence W Row- ley. Optimizing oblique projections for nonlinear systems using trajectories.SIAM Journal on Scientific Computing, 44(3):A1681–A1702, 2022

    2022

  43. [43]

    Model reduction for nonlinear systems by balanced trun- cation of state and gradient covariance.SIAM Journal on Scientific Computing, 45(5):A2325–A2355, 2023

    Samuel E Otto, Alberto Padovan, and Clarence W Rowley. Model reduction for nonlinear systems by balanced trun- cation of state and gradient covariance.SIAM Journal on Scientific Computing, 45(5):A2325–A2355, 2023

    2023

  44. [44]

    H Babaee and TP Sapsis. A minimization principle for the description of modes associated with finite-time instabili- ties.Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 472(2186), 2016

    2016

  45. [45]

    Babaee, M

    H. Babaee, M. Farazmand, G. Haller, and T. P . Sapsis. Reduced-order description of transient instabilities and computation of finite-time lyapunov exponents.Chaos, 27(6), 2017

    2017

  46. [46]

    Dynamics-informed deep learning for predicting ex- treme events.arXiv preprint arXiv:2603.10777, 2026

    Eirini Katsidoniotaki and Themistoklis P Sapsis. Dynamics-informed deep learning for predicting ex- treme events.arXiv preprint arXiv:2603.10777, 2026

    work page arXiv 2026

  47. [47]

    Spec- tral proper orthogonal decomposition and its relationship to dynamic mode decomposition and resolvent analysis

    Aaron Towne, Oliver T Schmidt, and Tim Colonius. Spec- tral proper orthogonal decomposition and its relationship to dynamic mode decomposition and resolvent analysis. Journal of Fluid Mechanics, 847:821–867, 2018

    2018

  48. [48]

    A conditional space– time pod formalism for intermittent and rare events: ex- ample of acoustic bursts in turbulent jets.Journal of Fluid Mechanics, 867:R2, 2019

    Oliver T Schmidt and Peter J Schmid. A conditional space– time pod formalism for intermittent and rare events: ex- ample of acoustic bursts in turbulent jets.Journal of Fluid Mechanics, 867:R2, 2019

    2019

  49. [49]

    Oliver T Schmidt. Data-driven forecasting of high- dimensional transient and stationary processes via space– time projection.Proceedings of the Royal Society A: Mathe- matical, Physical and Engineering Sciences, 482(2329), 2026

    2026

  50. [50]

    S. L. Brunton and J. N. Kutz.Data-Driven Science and Engi- neering: Machine Learning, Dynamical Systems, and Control. Cambridge University Press, 2nd edition, 2022

    2022

  51. [51]

    Sirovich

    L. Sirovich. Turbulence and the dynamics of coherent structures, parts I-III.Q. Appl. Math., XLV(3):561–590, 1987

    1987

  52. [52]

    B. Moore. Principal component analysis in linear systems: Controllability, observability, and model reduction.IEEE Transactions on Automatic Control, 26(1):17–32, 1981

    1981

  53. [53]

    Balancing for nonlinear systems

    Jacquelien MA Scherpen. Balancing for nonlinear systems. Systems & Control Letters, 21(2):143–153, 1993

    1993

  54. [54]

    Computation of nonlinear balanced realization and model reduction based on taylor series expansion.Systems & Control Letters, 57(4):283–289, 2008

    Kenji Fujimoto and Daisuke Tsubakino. Computation of nonlinear balanced realization and model reduction based on taylor series expansion.Systems & Control Letters, 57(4):283–289, 2008

    2008

  55. [55]

    SIAM, 2015

    Paul G Constantine.Active subspaces: Emerging ideas for dimension reduction in parameter studies. SIAM, 2015

    2015

  56. [56]

    Gradient-based dimension reduction of multivariate vector-valued functions.SIAM Journal on Scientific Computing, 42(1):A534–A558, 2020

    Olivier Zahm, Paul G Constantine, Cl ´ementine Prieur, and Youssef M Marzouk. Gradient-based dimension reduction of multivariate vector-valued functions.SIAM Journal on Scientific Computing, 42(1):A534–A558, 2020

    2020

  57. [57]

    Bradbury, R

    J. Bradbury, R. Frostig, P . Hawkins, M. J. Johnson, Y. Katariya, C. Leary, D. Maclaurin, G. Necula, A. Paszke, J. VanderPlas, S. Wanderman-Milne, and Q. Zhang. JAX: composable transformations of Python+NumPy pro- grams, 2018

    2018

  58. [58]

    Chaos control using least-squares support vector machines.International 13 journal of circuit theory and applications, 27(6):605–615, 1999

    Johan AK Suykens and Joos Vandewalle. Chaos control using least-squares support vector machines.International 13 journal of circuit theory and applications, 27(6):605–615, 1999

    1999

  59. [59]

    An investigation of chaotic kolmogorov flows.Physics of Fluids A: Fluid Dynamics, 3(4):681–696, 1991

    Nathan Platt, Lawrence Sirovich, and Nessan Fitzmaurice. An investigation of chaotic kolmogorov flows.Physics of Fluids A: Fluid Dynamics, 3(4):681–696, 1991

    1991

  60. [60]

    Numerical study of three-dimensional kolmogorov flow at high reynolds numbers.Journal of Fluid Mechanics, 306:293–323, 1996

    Vadim Borue and Steven A Orszag. Numerical study of three-dimensional kolmogorov flow at high reynolds numbers.Journal of Fluid Mechanics, 306:293–323, 1996

    1996

  61. [61]

    Clean numerical simulation of three-dimensional turbulent kolmogorov flow.Physics of Fluids, 37(10), 2025

    Shijie Qin and Shijun Liao. Clean numerical simulation of three-dimensional turbulent kolmogorov flow.Physics of Fluids, 37(10), 2025

    2025

  62. [62]

    Kolmogorov flow and laboratory simula- tion of it.Russ

    AM Obukhov. Kolmogorov flow and laboratory simula- tion of it.Russ. Math. Surv, 38(4):113–126, 1983

    1983

  63. [63]

    Closed-loop adaptive control of extreme events in a tur- bulent flow.Physical Review E, 100(3):033110, 2019

    Mohammad Farazmand and Themistoklis P Sapsis. Closed-loop adaptive control of extreme events in a tur- bulent flow.Physical Review E, 100(3):033110, 2019

    2019

  64. [64]

    Regenerative memory in time-delayed neuromorphic photonic res- onators.Scientific reports, 6(1):19510, 2016

    Bruno Romeira, Ricardo Av ´o, Jos ´e ML Figueiredo, St´ephane Barland, and Julien Javaloyes. Regenerative memory in time-delayed neuromorphic photonic res- onators.Scientific reports, 6(1):19510, 2016

    2016

  65. [65]

    Gerster, R

    M. Gerster, R. Berner, J. Sawicki, A. Zakharova, A. ˇSkoch, J. Hlinka, K. Lehnertz, and E. Sch ¨oll. Fitzhugh–nagumo oscillators on complex networks mimic epileptic-seizure- related synchronization phenomena.Chaos, 30(12), 2020

    2020

  66. [66]

    M. P . Nash and A. V . Panfilov. Electromechanical model of excitable tissue to study reentrant cardiac arrhythmias. Progress in biophysics and molecular biology, 85(2-3):501–522, 2004

    2004

  67. [67]

    Hubs, diversity, and synchronization in fitzhugh-nagumo oscillator networks: Resonance ef- fects and biophysical implications.Physical Review E, 103(5):052211, 2021

    Stefano Scialla, Alessandro Loppini, Marco Patriarca, and Els Heinsalu. Hubs, diversity, and synchronization in fitzhugh-nagumo oscillator networks: Resonance ef- fects and biophysical implications.Physical Review E, 103(5):052211, 2021

    2021

  68. [68]

    Resonance tongues and patterns in period- ically forced reaction-diffusion systems.Physical Re- view E—Statistical, Nonlinear, and Soft Matter Physics, 69(6):066217, 2004

    Anna L Lin, Aric Hagberg, Ehud Meron, and Harry L Swinney. Resonance tongues and patterns in period- ically forced reaction-diffusion systems.Physical Re- view E—Statistical, Nonlinear, and Soft Matter Physics, 69(6):066217, 2004

    2004

  69. [69]

    Cebri ´an-Lacasa, P

    D. Cebri ´an-Lacasa, P . Parra-Rivas, D. Ruiz-Reyn ´es, and L. Gelens. Six decades of the fitzhugh–nagumo model: A guide through its spatio-temporal dynamics and influence across disciplines.Physics Reports, 1096:1–39, 2024

    2024

  70. [70]

    Karnatak, G

    R. Karnatak, G. Ansmann, U. Feudel, and K. Lehnertz. Route to extreme events in excitable systems.Phys. Rev. E, 90(2):022917, 2014

    2014

  71. [71]

    Dysthe, H.E

    K. Dysthe, H.E. Krogstad, and P . M ¨uller. Oceanic rogue waves.Annu. Rev. Fluid Mech, 40(1):287–310, 2008

    2008

  72. [72]

    Optical rogue waves.nature, 450(7172):1054–1057, 2007

    Daniel R Solli, Claus Ropers, Prakash Koonath, and Bahram Jalali. Optical rogue waves.nature, 450(7172):1054–1057, 2007

    2007

  73. [73]

    Cousins and T.P

    W. Cousins and T.P . Sapsis. Reduced-order precursors of rare events in unidirectional nonlinearwaterwaves.J. Fluid Mech, 790(3):368–388, 2016

    2016

  74. [74]

    Cousins and T.P

    W. Cousins and T.P . Sapsis. Unsteady evolution of local- ized unidirectional deep-water wave groups.Phys. Rev. E, 91(6):063204, 2015

    2015

  75. [75]

    Modulational instability and non-gaussian statistics in experimental random water-wave trains

    Miguel Onorato, Alfred Richard Osborne, M Serio, and L Cavaleri. Modulational instability and non-gaussian statistics in experimental random water-wave trains. Physics of Fluids, 17(7), 2005

    2005

  76. [76]

    A mathematical guide to operator learning

    Nicolas Boull ´e and Alex Townsend. A mathematical guide to operator learning. InHandbook of Numerical Analysis, volume 25, pages 83–125. Elsevier, 2024

    2024

  77. [77]

    Operator learning without the adjoint.Journal of Machine Learning Research, 25(364):1–54, 2024

    Nicolas Boull ´e, Diana Halikias, Samuel E Otto, and Alex Townsend. Operator learning without the adjoint.Journal of Machine Learning Research, 25(364):1–54, 2024

    2024

  78. [78]

    Z. Li, N. Borislavov Kovachki, K. Azizzadenesheli, B. Liu, K. Bhattacharya, A. Stuart, and A. Anandkumar. Fourier neural operator for parametric partial differential equa- tions. InICLR, 2021

    2021

  79. [79]

    Fourth-order time-stepping for stiff pdes.SIAM Journal on Scientific Com- puting, 26(4):1214–1233, 2005

    Aly-Khan Kassam and Lloyd N Trefethen. Fourth-order time-stepping for stiff pdes.SIAM Journal on Scientific Com- puting, 26(4):1214–1233, 2005

    2005

  80. [80]

    Machine learning– accelerated computational fluid dynamics.Proceedings of the National Academy of Sciences, 118(21):e2101784118, 2021

    Dmitrii Kochkov, Jamie A Smith, Ayya Alieva, Qing Wang, Michael P Brenner, and Stephan Hoyer. Machine learning– accelerated computational fluid dynamics.Proceedings of the National Academy of Sciences, 118(21):e2101784118, 2021

    2021

Showing first 80 references.