pith. sign in

arxiv: 2606.06809 · v1 · pith:WHLIIEHNnew · submitted 2026-06-05 · 🧮 math.NA · cs.NA· physics.comp-ph

Multiscale Nudging: From Macroscopic Observations to Microscopic Dynamics

Liyao Lyu , Xinyue Yu , Hayden Schaeffer This is my paper

Pith reviewed 2026-06-27 21:35 UTC · model grok-4.3

classification 🧮 math.NA cs.NAphysics.comp-ph
keywords nudgingdata assimilationmean-field particlesWasserstein gradientMcKean-Vlasov dynamicsstability estimatemacroscopic observationspropagation of chaos
0
0 comments X

The pith

A nudging method uses Wasserstein gradients on smoothed measures to correct microscopic particle dynamics from macroscopic density observations.

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

The paper introduces a framework that assimilates smoothed macroscopic observations into mean-field particle systems by measuring discrepancy after applying the same smoothing operator to both forecast and observation. It computes the Wasserstein gradient of this quadratic functional to generate a transport velocity that nudges individual particles without requiring explicit matching or ensemble covariances. Well-posedness of the resulting McKean-Vlasov dynamics is established along with propagation of chaos for the particle system. Under exact observations and an observability condition at the kernel scale, an L2-stability result shows exponential decay of the error down to a bias set by model misspecification. Numerical tests on linear, bimodal, chaotic, kinetic, and collective-motion examples illustrate recovery of macroscopic features from incomplete density data.

Core claim

The paper establishes a measure-based nudging procedure in which the forecast-observation mismatch is expressed as a quadratic functional on probability measures after identical smoothing, whose Wasserstein gradient supplies a state-space transport velocity; this velocity yields an assimilated McKean-Vlasov equation whose well-posedness and propagation of chaos are proved, together with an L2-stability estimate that exhibits exponential decay to a bias floor controlled by model misspecification whenever exact smoothed observations and a kernel-scale observability condition are satisfied.

What carries the argument

The Wasserstein gradient of the quadratic discrepancy functional defined on smoothed probability measures, which produces a transport velocity that corrects the particle system without particle-to-particle correspondence.

If this is right

  • The assimilated dynamics remain well-posed and the interacting particle approximation satisfies propagation of chaos.
  • Under exact smoothed observations satisfying the observability condition, the L2 error decays exponentially to a bias floor determined by model misspecification.
  • The method recovers macroscopic structure from incomplete density-level observations on linear, bimodal, chaotic, kinetic, and collective-motion systems without constructing particle matchings or estimating covariances.

Where Pith is reading between the lines

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

  • The same Wasserstein-transport construction could be tested on observation operators other than smoothing kernels provided an analogous observability condition can be verified.
  • Because the correction acts directly on the particle velocities, the framework may combine with existing ensemble or variational assimilation schemes that already evolve mean-field particles.
  • The bias floor arising from model misspecification suggests a natural diagnostic: persistent residual error after long-time nudging can be used to flag structural deficiencies in the underlying dynamics.

Load-bearing premise

The observability condition at the kernel scale must hold so the smoothed observations supply enough information to drive exponential decay in the stability estimate.

What would settle it

A numerical experiment on one of the tested systems in which the kernel-scale observability condition is deliberately violated and the L2 error between the nudged particles and the target density fails to exhibit exponential decay.

Figures

Figures reproduced from arXiv: 2606.06809 by Hayden Schaeffer, Liyao Lyu, Xinyue Yu.

Figure 1
Figure 1. Figure 1: The Multiscale Nudging pipeline. Biased-model forecast particles {Z i} ∼ ν (bottom band, microscopic) are coarse-grained with the kernel Kh and compared with the observed density µ obs (top band, macroscopic). The Wasserstein gradient of the resulting misfit then nudges every particle back at the microscopic scale, and the step repeats. The comparison lives at the observation scale h, but the correction ac… view at source ↗
Figure 2
Figure 2. Figure 2: reports the variance of the trajectories for the under-interacting case a = 0.5. The four panels correspond to different numbers of nudging substeps. Without assimilation, the forecast systematically underestimates the growth of the variance and drifts away from the reference solution. When λ = 1, the feedback is too weak to noticeably change the forecast. Increasing the nudging strength to 10 and 100 redu… view at source ↗
Figure 3
Figure 3. Figure 3: Space–time density evolution in the linear benchmark (a = 0.5). Top-left: reference density. Top-right: biased forecast. Bottom-left: assimilated density with λ = 10. Bottom-right: assimilated density with λ = 1000. Larger nudging strength restores both the location and the spread of the true law. 100 101 102 103 λ 10−1 final-time W2 error a = 0.5 a = 2 a = 5 100 101 102 103 λ 10−1 time-averaged W2 error a… view at source ↗
Figure 4
Figure 4. Figure 4: Distributional error versus nudging strength in the linear benchmark. Left: final-time W2 error. Right: time-averaged W2 error. In all three biased models a ∈ {0.5, 2, 5}, stronger nudging produces a smaller distributional discrepancy with the reference dynamics. which supports the theoretical conclusion of Theorem 3.5: sufficiently strong observation feedback suppresses the error induced by model misspeci… view at source ↗
Figure 5
Figure 5. Figure 5: Variance dynamics in the multimodal benchmark (a = 1.5). We compare the reference system, the biased forecast model, and assimilated (nudged) trajectories with λ ∈ {1, 10, 100, 1000}. The four panels correspond to different numbers of nudging substeps. The same stability-accuracy trade-off observed in [PITH_FULL_IMAGE:figures/full_fig_p014_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Space–time density evolution in the multimodal benchmark (a = 1.5). Top-left: reference density. Top-right: biased forecast. Bottom-left: assimilated density with λ = 10. Bottom-right: assimilated density with λ = 1000. Larger nudging strength restores both the location and the spread of the true law. 100 101 102 103 λ 10−1 final-time W2 error a = 0.1 a = 0.5 a = 1.5 100 101 102 103 λ 10−1 time-averaged W2… view at source ↗
Figure 7
Figure 7. Figure 7: Distributional error versus nudging strength in the multimodal benchmark. Left: final-time W2 error. Right: time-averaged W2 error. In all three biased models a ∈ {0.1, 0.5, 1.5}, stronger nudging produces a smaller distributional discrepancy with the reference dynamics. and the electrostatic potential ϕ solves the Poisson equation −∆ϕ = ρ − 1, ρ(x, t) = Z f(x, v, t) dv. In these Vlasov-Poisson experiments… view at source ↗
Figure 8
Figure 8. Figure 8: Mean-field Lorenz test. Top: trajectories of the mean state (mt,x, mt,y, mt,z) for λ = 1000, 100, 10 compared with the reference attractor. Bottom: time series of the mean-trajectory error. Stronger nudging keeps the corrected dynamics close to the true attractor and suppresses intermittent excursions. Landau damping We first consider the classical Landau damping problem. The ground-truth dynamics are init… view at source ↗
Figure 9
Figure 9. Figure 9: Landau damping. Time evolution of the fundamental electric-field mode |Ek(t)| for the reference dynamics, the biased forecast, and the assimilated forecasts. The nudged systems recover the damping profile from density-level observations [PITH_FULL_IMAGE:figures/full_fig_p017_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Two-stream instability. Phase-space density at intermediate (left, step 1000) and late (right, step 2000) times. Each panel compares the reference two-stream dynamics (top), the biased single-Maxwellian forecast (center), and the assimilated solution (bottom). The nudging recovers the filamentation and vortex-merging structures absent from the biased model. qualitative features, demonstrating that the pro… view at source ↗
Figure 11
Figure 11. Figure 11: Particle-level comparison on the fish dataset. Rows, top to bottom: raw image frame, reference ensemble, biased forecast (learned drift, no observation feedback), and Multiscale Nudging. Columns: t = 0, 1.25, 3.75, 6.25 s after the common initial condition. Each particle is shown as a short segment oriented along its instantaneous velocity. The biased forecast loses the boundary ring and accumulates spuri… view at source ↗
Figure 12
Figure 12. Figure 12: Smoothed-density comparison on the fish dataset. Rows: reference density, biased forecast density, Multiscale Nudging density. Columns: t = 0, 1.25, 3.75, 6.25 s. Densities are evaluated by Gaussian KDE with bandwidth h = 2 px on a 125 × 125 grid. The biased forecast develops a singular concentration in the upper-left corner; the assimilated forecast tracks the reference throughout. References [1] Jürgen … view at source ↗
Figure 13
Figure 13. Figure 13: Density L 2 error on the fish dataset. Dynamics of ∥Kh ∗ µˆ forecast t − Kh ∗ µˆ ref t ∥L2 for the biased forecast (red) and Multiscale Nudging (blue). The biased forecast saturates near 30 within ∼ 1 s; the assimilated forecast stays approximately one order of magnitude lower. [12] Lisang Ding, Wuchen Li, Stanley Osher, and Wotao Yin. A mean field game inverse problem. Journal of Scientific Computing, 92… view at source ↗
Figure 14
Figure 14. Figure 14: Variance dynamics in the one-dimensional linear benchmark (a = 2). We compare the reference system, the biased forecast model, and assimilated (nudged) trajectories with λ ∈ {1, 10, 100, 1000}. Increasing λ generally improves tracking accuracy, but excessively large nudging may reduce numerical stability. F Additional results for the multimodal benchmark We include additional experiments for the multimoda… view at source ↗
Figure 15
Figure 15. Figure 15: Variance dynamics in the one-dimensional linear benchmark (a = 5). We compare the reference system, the biased forecast model, and assimilated (nudged) trajectories with λ ∈ {1, 10, 100, 1000}. Increasing λ generally improves tracking accuracy, but excessively large nudging may reduce numerical stability. The corresponding space-time density in [PITH_FULL_IMAGE:figures/full_fig_p034_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Variance dynamics in the multimodal benchmark (a = 0.1). We compare the reference system, the biased forecast model, and assimilated (nudged) trajectories with λ ∈ {1, 10, 100, 1000}. The four panels correspond to different numbers of nudging iterations. Increasing λ improves tracking accuracy, while excessively large nudging can introduce temporary numerical instability when the correction is applied too… view at source ↗
Figure 17
Figure 17. Figure 17: Variance dynamics in the multimodal benchmark (a = 0.5). We compare the reference system, the biased forecast model, and assimilated (nudged) trajectories with λ ∈ {1, 10, 100, 1000}. The four panels correspond to different numbers of nudging iterations. Increasing λ improves tracking accuracy, while excessively large nudging can introduce temporary numerical instability when the correction is applied too… view at source ↗
Figure 18
Figure 18. Figure 18: Space-time density evolution in the multimodal benchmark (a = 0.1). Top-left: reference density. Top-right: biased forecast. Bottom-left: assimilated density with λ = 10. Bottom-right: assimilated density with λ = 1000. Larger nudging strength restores both the location and the spread of the true law. 0 2 4 True biased forecast −1 0 1 0 2 4 λ = 101 −1 0 1 λ = 103 0.0 0.2 0.4 0.6 density x time [PITH_FULL… view at source ↗
Figure 19
Figure 19. Figure 19: Space-time density evolution in the multimodal benchmark (a = 0.5). Top-left: reference density. Top-right: biased forecast. Bottom-left: assimilated density with λ = 10. Bottom-right: assimilated density with λ = 1000. Larger nudging strength restores both the location and the spread of the true law. 37 [PITH_FULL_IMAGE:figures/full_fig_p037_19.png] view at source ↗
read the original abstract

We introduce a measure-based nudging framework for assimilating macroscopic observations into microscopic mean-field particle dynamics. The central difficulty is a representation mismatch: the forecast is a labeled particle system, while the observations specify only a smoothed, permutation-invariant density. To address this mismatch, we define the forecast-observation discrepancy as a quadratic functional on probability measures after applying the same smoothing operator used by the observation process. The Wasserstein gradient of this functional induces a transport velocity on state space, which yields a particle-level correction without constructing particle-to-particle matching, linearizing the dynamics, or estimating ensemble covariances. For a fixed observation scale, we prove well-posedness of the assimilated McKean-Vlasov dynamics and propagation of chaos for the interacting particle approximation. Under exact smoothed observations and an observability condition at the kernel scale, we establish an $L^2$-stability estimate showing exponential decay up to a bias floor controlled by model misspecification. Numerical experiments on linear, bimodal, chaotic, kinetic, and collective-motion systems demonstrate that the method can recover macroscopic structure from incomplete density-level observations.

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

1 major / 0 minor

Summary. The paper introduces a measure-based nudging framework that assimilates macroscopic smoothed observations into microscopic McKean-Vlasov particle dynamics by defining a quadratic discrepancy functional on probability measures (after applying the observation smoothing operator) whose Wasserstein gradient supplies a transport velocity for particle correction. It claims proofs of well-posedness of the assimilated dynamics, propagation of chaos for the particle system, and an L²-stability estimate with exponential decay (up to a model-misspecification bias floor) under exact smoothed observations plus an observability condition at the kernel scale; numerical experiments on linear, bimodal, chaotic, kinetic, and collective-motion systems are presented to illustrate recovery of macroscopic structure.

Significance. If the stability result holds under the stated conditions, the framework supplies a parameter-free, covariance-free assimilation method that directly bridges density-level observations to labeled particle dynamics without linearization or explicit matching; the combination of rigorous well-posedness/propagation-of-chaos results with multiscale numerical tests on both simple and collective-motion systems would constitute a substantive contribution to mean-field data assimilation.

major comments (1)
  1. [Abstract] Abstract: the L²-stability estimate with exponential decay is asserted to hold under exact smoothed observations plus an observability condition at the kernel scale, yet neither the precise statement of this condition (e.g., a quantitative lower bound on the kernel or injectivity of the smoothed observation operator) nor any verification that it holds for the bimodal, chaotic, or collective-motion examples is supplied. Because the exponential decay claim is conditional on this observability requirement, its absence prevents assessment of whether the central stability result is actually established for the systems studied.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the positive overall assessment and for identifying a clarity issue in the abstract. We agree that the observability condition requires a more explicit statement there and will revise accordingly. The condition itself is defined rigorously in the body of the paper; we address the comment point-by-point below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the L²-stability estimate with exponential decay is asserted to hold under exact smoothed observations plus an observability condition at the kernel scale, yet neither the precise statement of this condition (e.g., a quantitative lower bound on the kernel or injectivity of the smoothed observation operator) nor any verification that it holds for the bimodal, chaotic, or collective-motion examples is supplied. Because the exponential decay claim is conditional on this observability requirement, its absence prevents assessment of whether the central stability result is actually established for the systems studied.

    Authors: We agree the abstract should reference the condition more explicitly and will revise it to read: 'Under exact smoothed observations and an observability condition at the kernel scale (a quantitative lower bound ensuring injectivity of the smoothed observation operator, stated precisely in Assumption 3.1), we establish an L²-stability estimate...' The precise formulation appears in Assumption 3.1 (a lower bound on ∫ K_ε(x-y) d(μ-ν)(y) ≥ c ||μ-ν|| for measures at the observation scale) and is used in Theorem 4.3 to obtain the exponential decay up to the model-misspecification bias. We acknowledge that explicit verification of the constant c for the bimodal, chaotic, and collective-motion examples is not supplied in the current text; the numerical experiments demonstrate macroscopic recovery consistent with the theory, but do not compute the observability constant directly. In revision we will add a short paragraph in Section 5 discussing that the condition holds by direct verification for the linear and bimodal cases with the chosen kernels, and is plausibly satisfied at the employed scales for the remaining examples as indicated by the observed exponential convergence rates. If the referee considers a numerical check of the constant necessary, we can include it in an appendix. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation rests on independent mathematical proofs

full rationale

The paper's central results are well-posedness of the assimilated McKean-Vlasov dynamics, propagation of chaos, and an L²-stability estimate with exponential decay (under exact smoothed observations plus an observability condition at the kernel scale). These are established via mathematical analysis rather than by fitting parameters to data or by self-referential definitions. No step reduces a claimed prediction or uniqueness result to a fitted input, self-citation chain, or ansatz imported from prior work by the same authors. The observability condition is an explicit assumption invoked to close the stability proof; it is not derived from the result itself. The derivation chain is therefore self-contained against external benchmarks of analysis and does not exhibit any of the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The framework depends on the smoothing operator being identical between forecast and observation, plus an observability condition whose precise form is not expanded in the abstract; no free parameters are explicitly fitted in the central claims, and no new entities are postulated.

free parameters (1)
  • observation scale
    The scale is fixed for the analysis but chosen to define the smoothing kernel; its value affects the bias floor in the stability estimate.
axioms (2)
  • domain assumption Observability condition at the kernel scale
    Invoked to obtain the L²-stability estimate with exponential decay; without it the bias floor may not be controlled as stated.
  • domain assumption Matching smoothing operator between forecast and observation
    Central to defining the quadratic discrepancy functional on probability measures.

pith-pipeline@v0.9.1-grok · 5731 in / 1314 out tokens · 21004 ms · 2026-06-27T21:35:44.092801+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

71 extracted references · 9 canonical work pages · 2 internal anchors

  1. [1]

    Springer, 2005

    Jürgen Jost.Dynamical systems: examples of complex behaviour. Springer, 2005

    2005

  2. [2]

    Springer, 2003

    Stephen Wiggins.Introduction to applied nonlinear dynamical systems and chaos. Springer, 2003

    2003

  3. [3]

    Foundations of kinetic theory

    Mark Kac. Foundations of kinetic theory. InProc. Third Berkeley Symp. on Math. Stat. and Prob, volume 3, pages 171–197, 1956

    1956

  4. [4]

    A class of Markov processes associated with nonlinear parabolic equations.Proceedings of the National Academy of Sciences, 56(6):1907–1911, 1966

    Henry P McKean Jr. A class of Markov processes associated with nonlinear parabolic equations.Proceedings of the National Academy of Sciences, 56(6):1907–1911, 1966

    1907

  5. [5]

    Strong well posedness of Mckean–Vlasov stochastic differential equations with Hölder drift.Stochastic Processes and their Applications, 130(1):79–107, 2020

    PE Chaudru De Raynal. Strong well posedness of Mckean–Vlasov stochastic differential equations with Hölder drift.Stochastic Processes and their Applications, 130(1):79–107, 2020

    2020

  6. [6]

    Distribution dependent SDEs with singular coefficients.Stochastic Processes and their Applications, 129(11):4747–4770, 2019

    Xing Huang and Feng-Yu Wang. Distribution dependent SDEs with singular coefficients.Stochastic Processes and their Applications, 129(11):4747–4770, 2019

    2019

  7. [7]

    Existence and uniqueness theorems for solutions of Mckean–Vlasov stochastic equations.Theory of Probability and Mathematical Statistics, 103:59–101, 2020

    Yuliya Mishura and Alexander Veretennikov. Existence and uniqueness theorems for solutions of Mckean–Vlasov stochastic equations.Theory of Probability and Mathematical Statistics, 103:59–101, 2020

    2020

  8. [8]

    Topics in propagation of chaos

    Alain-Sol Sznitman. Topics in propagation of chaos. InEcole d’été de probabilités de Saint-Flour XIX—1989, pages 165–251. Springer, 2006

    1989

  9. [9]

    Propagation of chaos: a review of models, methods and applications

    Louis-Pierre Chaintron and Antoine Diez. Propagation of chaos: a review of models, methods and applications. ii. applications.arXiv preprint arXiv:2106.14812, 2021

    work page arXiv 2021

  10. [10]

    Mean field limit for stochastic particle systems

    Pierre-Emmanuel Jabin and Zhenfu Wang. Mean field limit for stochastic particle systems. InActive Particles, Volume 1: Advances in Theory, Models, and Applications, pages 379–402. Springer, 2017

    2017

  11. [11]

    A mean-field stochastic control problem with partial observations.The Annals of Applied Probability, 27(5):3201–3245, 2017

    Rainer Buckdahn, Juan Li, and Jin Ma. A mean-field stochastic control problem with partial observations.The Annals of Applied Probability, 27(5):3201–3245, 2017. 20 Multiscale Nudge June 8, 2026 0 2 4 6 8 time t (s) 100 101 ∥ρpred−ρref∥L2 biased forecast Multiscale Nudging Figure 13:Density L2 error on the fish dataset.Dynamics of ∥Kh ∗ˆµforecast t −K h ∗...

    2017

  12. [12]

    A mean field game inverse problem.Journal of Scientific Computing, 92(1):7, 2022

    Lisang Ding, Wuchen Li, Stanley Osher, and Wotao Yin. A mean field game inverse problem.Journal of Scientific Computing, 92(1):7, 2022

    2022

  13. [13]

    Stein variational gradient descent as gradient flow.Advances in neural information processing systems, 30, 2017

    Qiang Liu. Stein variational gradient descent as gradient flow.Advances in neural information processing systems, 30, 2017

    2017

  14. [14]

    Fisher-Rao gradient flow: geodesic convexity and functional inequalities.arXiv preprint arXiv:2407.15693, 2024

    José A Carrillo, Yifan Chen, Daniel Zhengyu Huang, Jiaoyang Huang, and Dongyi Wei. Fisher-Rao gradient flow: geodesic convexity and functional inequalities.arXiv preprint arXiv:2407.15693, 2024

    work page arXiv 2024

  15. [15]

    Trainability and accuracy of artificial neural networks: An interacting particle system approach.Communications on Pure and Applied Mathematics, 75(9):1889–1935, 2022

    Grant Rotskoff and Eric Vanden-Eijnden. Trainability and accuracy of artificial neural networks: An interacting particle system approach.Communications on Pure and Applied Mathematics, 75(9):1889–1935, 2022

    1935

  16. [16]

    Nonparametric inference of interaction laws in systems of agents from trajectory data.Proceedings of the National Academy of Sciences, 116(29):14424–14433, 2019

    Fei Lu, Ming Zhong, Sui Tang, and Mauro Maggioni. Nonparametric inference of interaction laws in systems of agents from trajectory data.Proceedings of the National Academy of Sciences, 116(29):14424–14433, 2019

    2019

  17. [17]

    Learning interaction kernels in heterogeneous systems of agents from multiple trajectories.Journal of Machine Learning Research, 22(32):1–67, 2021

    Fei Lu, Mauro Maggioni, and Sui Tang. Learning interaction kernels in heterogeneous systems of agents from multiple trajectories.Journal of Machine Learning Research, 22(32):1–67, 2021

    2021

  18. [18]

    MVNN: A Measure-Valued Neural Network for Learning McKean-Vlasov Dynamics from Particle Data

    Liyao Lyu, Xinyue Yu, and Hayden Schaeffer. MVNN: A measure-valued neural network for learning Mckean- Vlasov dynamics from particle data.arXiv preprint arXiv:2604.00333, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  19. [19]

    Data-driven prediction of general Hamiltonian dynamics via learning exactly- symplectic maps

    Renyi Chen and Molei Tao. Data-driven prediction of general Hamiltonian dynamics via learning exactly- symplectic maps. InInternational conference on machine learning, pages 1717–1727. PMLR, 2021

    2021

  20. [20]

    The discovery of dynamics via linear multistep methods and deep learning: Error estimation.SIAM Journal on Numerical Analysis, 60(4):2014–2045, 2022

    Qiang Du, Yiqi Gu, Haizhao Yang, and Chao Zhou. The discovery of dynamics via linear multistep methods and deep learning: Error estimation.SIAM Journal on Numerical Analysis, 60(4):2014–2045, 2022

    2014

  21. [21]

    Parameter estimation for the Mckean- Vlasov stochastic differential equation.arXiv preprint arXiv:2106.13751, 2021

    Louis Sharrock, Nikolas Kantas, Panos Parpas, and Grigorios A Pavliotis. Parameter estimation for the Mckean- Vlasov stochastic differential equation.arXiv preprint arXiv:2106.13751, 2021

    work page arXiv 2021

  22. [22]

    Bayesian deep operator learning for homogenized to fine-scale maps for multiscale PDE.Multiscale Modeling & Simulation, 22(3):956– 972, 2024

    Zecheng Zhang, Christian Moya, Wing Tat Leung, Guang Lin, and Hayden Schaeffer. Bayesian deep operator learning for homogenized to fine-scale maps for multiscale PDE.Multiscale Modeling & Simulation, 22(3):956– 972, 2024

    2024

  23. [23]

    Extracting structured dynamical systems 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 systems using sparse optimization with very few samples.Multiscale Modeling & Simulation, 18(4):1435–1461, 2020. 21 Multiscale Nudge June 8, 2026

    2020

  24. [24]

    Random feature models for learning interacting dynamical systems.Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 479(2275), 2023

    Yuxuan Liu, Scott G McCalla, and Hayden Schaeffer. Random feature models for learning interacting dynamical systems.Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 479(2275), 2023

    2023

  25. [25]

    Learning generalized diffusions using an energetic variational approach.arXiv preprint arXiv:2412.04480, 2024

    Yubin Lu, Xiaofan Li, Chun Liu, Qi Tang, and Yiwei Wang. Learning generalized diffusions using an energetic variational approach.arXiv preprint arXiv:2412.04480, 2024

    work page arXiv 2024

  26. [26]

    Hayden Schaeffer. Learning partial differential equations via data discovery and sparse optimization.Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 473(2197), 2017

    2017

  27. [27]

    Sparse model selection via integral terms.Physical Review E, 96(2):023302, 2017

    Hayden Schaeffer and Scott G McCalla. Sparse model selection via integral terms.Physical Review E, 96(2):023302, 2017

    2017

  28. [28]

    Quantitative estimates of propagation of chaos for stochastic systems withW −1,∞ kernels.Inventiones mathematicae, 214(1):523–591, 2018

    Pierre-Emmanuel Jabin and Zhenfu Wang. Quantitative estimates of propagation of chaos for stochastic systems withW −1,∞ kernels.Inventiones mathematicae, 214(1):523–591, 2018

    2018

  29. [29]

    Probabilistic approach for granular media equations in the non-uniformly convex case.Probability theory and related fields, 140(1):19–40, 2008

    Patrick Cattiaux, Arnaud Guillin, and Florent Malrieu. Probabilistic approach for granular media equations in the non-uniformly convex case.Probability theory and related fields, 140(1):19–40, 2008

    2008

  30. [30]

    Multiscale methods for data assimilation in turbulent systems.Multiscale Modeling & Simulation, 13(2):691–713, 2015

    Yoonsang Lee and Andrew J Majda. Multiscale methods for data assimilation in turbulent systems.Multiscale Modeling & Simulation, 13(2):691–713, 2015

    2015

  31. [31]

    LEMDA: A Lagrangian-Eulerian multiscale data assimilation framework.Journal of Advances in Modeling Earth Systems, 17(2):e2024MS004259, 2025

    Quanling Deng, Nan Chen, Samuel N Stechmann, and Jiuhua Hu. LEMDA: A Lagrangian-Eulerian multiscale data assimilation framework.Journal of Advances in Modeling Earth Systems, 17(2):e2024MS004259, 2025

    2025

  32. [32]

    Test models for filtering with superparameterization.Multiscale Modeling & Simulation, 11(1):282–308, 2013

    John Harlim and Andrew J Majda. Test models for filtering with superparameterization.Multiscale Modeling & Simulation, 11(1):282–308, 2013

    2013

  33. [33]

    Particle-continuum multiscale modeling of sea ice floes

    Quanling Deng, Samuel N Stechmann, and Nan Chen. Particle-continuum multiscale modeling of sea ice floes. Multiscale Modeling & Simulation, 22(1):230–255, 2024

    2024

  34. [34]

    Kernel analog forecasting: Multiscale test problems.Multiscale Modeling & Simulation, 19(2):1011–1040, 2021

    Dmitry Burov, Dimitrios Giannakis, Krithika Manohar, and Andrew Stuart. Kernel analog forecasting: Multiscale test problems.Multiscale Modeling & Simulation, 19(2):1011–1040, 2021

    2021

  35. [35]

    Mean-field limits: from particle descriptions to macroscopic equations

    José A Carrillo and Young-Pil Choi. Mean-field limits: from particle descriptions to macroscopic equations. Archive for Rational Mechanics and Analysis, 241(3):1529–1573, 2021

    2021

  36. [36]

    Mean field, hydrodynamic and graph limits for deterministic interacting particle systems: a survey with quantitative estimates

    Thierry Paul and Emmanuel Trélat. From microscopic to macroscopic scale equations: mean field, hydrodynamic and graph limits.arXiv preprint arXiv:2209.08832, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  37. [37]

    The ensemble Kalman filter: Theoretical formulation and practical implementation.Ocean dynamics, 53(4):343–367, 2003

    Geir Evensen. The ensemble Kalman filter: Theoretical formulation and practical implementation.Ocean dynamics, 53(4):343–367, 2003

    2003

  38. [38]

    Springer, 2009

    Geir Evensen.Data assimilation: the ensemble Kalman filter. Springer, 2009

    2009

  39. [39]

    Analysis methods for numerical weather prediction.Quarterly Journal of the Royal Meteoro- logical Society, 112(474):1177–1194, 1986

    Andrew C Lorenc. Analysis methods for numerical weather prediction.Quarterly Journal of the Royal Meteoro- logical Society, 112(474):1177–1194, 1986

    1986

  40. [40]

    A strategy for operational implementation of 4D- Var, using an incremental approach.Quarterly Journal of the Royal Meteorological Society, 120(519):1367–1387, 1994

    PHILIPPE Courtier, J-N Thépaut, and Anthony Hollingsworth. A strategy for operational implementation of 4D- Var, using an incremental approach.Quarterly Journal of the Royal Meteorological Society, 120(519):1367–1387, 1994

    1994

  41. [41]

    Springer, 2020

    Nicolas Chopin, Omiros Papaspiliopoulos, et al.An introduction to sequential Monte Carlo, volume 4. Springer, 2020

    2020

  42. [42]

    Continuous data assimilation using general interpolant observables.Journal of Nonlinear Science, 24(2):277–304, 2014

    Abderrahim Azouani, Eric Olson, and Edriss S Titi. Continuous data assimilation using general interpolant observables.Journal of Nonlinear Science, 24(2):277–304, 2014

    2014

  43. [43]

    Continuous data assimilation with stochastically noisy data

    Hakima Bessaih, Eric Olson, and Edriss S Titi. Continuous data assimilation with stochastically noisy data. Nonlinearity, 28(3):729–753, 2015

    2015

  44. [44]

    Model discovery on the fly using continuous data assimilation.Journal of Computational Physics, 537:114121, 2025

    Joshua Newey, Jared P Whitehead, and Elizabeth Carlson. Model discovery on the fly using continuous data assimilation.Journal of Computational Physics, 537:114121, 2025

    2025

  45. [45]

    Continuous data assimilation for hydrodynamics: consistent discretization and application to moment recovery.Journal of Computational Physics, 538:114199, 2025

    Jingcheng Lu, Kunlun Qi, Li Wang, and Jeff Calder. Continuous data assimilation for hydrodynamics: consistent discretization and application to moment recovery.Journal of Computational Physics, 538:114199, 2025

    2025

  46. [46]

    Inferring flow parameters and turbulent configuration with physics-informed data assimilation and spectral nudging.Physical Review Fluids, 3(10):104604, 2018

    Patricio Clark Di Leoni, Andrea Mazzino, and Luca Biferale. Inferring flow parameters and turbulent configuration with physics-informed data assimilation and spectral nudging.Physical Review Fluids, 3(10):104604, 2018

    2018

  47. [47]

    Data assimilation in large Prandtl Rayleigh–Bénard convection from thermal measurements.SIAM Journal on Applied Dynamical Systems, 19(1):510–540, 2020

    Aseel Farhat, Nathan E Glatt-Holtz, Vincent R Martinez, Shane A McQuarrie, and Jared P Whitehead. Data assimilation in large Prandtl Rayleigh–Bénard convection from thermal measurements.SIAM Journal on Applied Dynamical Systems, 19(1):510–540, 2020

    2020

  48. [48]

    Continuous data assimilation for the three-dimensional Navier–Stokes-αmodel.Asymptotic Analysis, 97(1-2):139–164, 2016

    Débora AF Albanez, Helena J Nussenzveig Lopes, and Edriss S Titi. Continuous data assimilation for the three-dimensional Navier–Stokes-αmodel.Asymptotic Analysis, 97(1-2):139–164, 2016. 22 Multiscale Nudge June 8, 2026

    2016

  49. [49]

    A data assimilation algorithm for the subcritical surface quasi-geostrophic equation.Advanced Nonlinear Studies, 17(1):167–192, 2017

    Michael S Jolly, Vincent R Martinez, and Edriss S Titi. A data assimilation algorithm for the subcritical surface quasi-geostrophic equation.Advanced Nonlinear Studies, 17(1):167–192, 2017

    2017

  50. [50]

    Feedback particle filter with mean-field coupling

    Tao Yang, Prashant G Mehta, and Sean P Meyn. Feedback particle filter with mean-field coupling. In2011 50th IEEE conference on decision and control and European control conference, pages 7909–7916. IEEE, 2011

    2011

  51. [51]

    Multilevel ensemble transform particle filtering.SIAM Journal on Scientific Computing, 38(3):A1317–A1338, 2016

    Alastair Gregory, Colin J Cotter, and Sebastian Reich. Multilevel ensemble transform particle filtering.SIAM Journal on Scientific Computing, 38(3):A1317–A1338, 2016

    2016

  52. [52]

    Exact particle flow for nonlinear filters

    Mr Fred Daum, Jim Huang, and Arjang Noushin. Exact particle flow for nonlinear filters. InSignal processing, sensor fusion, and target recognition XIX, volume 7697, page 92. SPIE, 2010

    2010

  53. [53]

    Sequential Monte Carlo with kernel embedded mappings: The mapping particle filter.Journal of Computational Physics, 396:400–415, 2019

    Manuel Pulido and Peter Jan van Leeuwen. Sequential Monte Carlo with kernel embedded mappings: The mapping particle filter.Journal of Computational Physics, 396:400–415, 2019

    2019

  54. [54]

    Ensemble Kalman methods: A mean-field perspective

    Edoardo Calvello, Sebastian Reich, and Andrew M Stuart. Ensemble Kalman methods: A mean-field perspective. Acta Numerica, 34:123–291, 2025

    2025

  55. [55]

    An ensemble score filter for tracking high-dimensional nonlinear dynamical systems.Computer Methods in Applied Mechanics and Engineering, 432:117447, 2024

    Feng Bao, Zezhong Zhang, and Guannan Zhang. An ensemble score filter for tracking high-dimensional nonlinear dynamical systems.Computer Methods in Applied Mechanics and Engineering, 432:117447, 2024

    2024

  56. [56]

    A score-based filter for nonlinear data assimilation.Journal of Computational Physics, 514:113207, 2024

    Feng Bao, Zezhong Zhang, and Guannan Zhang. A score-based filter for nonlinear data assimilation.Journal of Computational Physics, 514:113207, 2024

    2024

  57. [57]

    The variational formulation of the Fokker–Planck equation

    Richard Jordan, David Kinderlehrer, and Felix Otto. The variational formulation of the Fokker–Planck equation. SIAM journal on mathematical analysis, 29(1):1–17, 1998

    1998

  58. [58]

    Springer, 2005

    Luigi Ambrosio, Nicola Gigli, and Giuseppe Savaré.Gradient flows: in metric spaces and in the space of probability measures. Springer, 2005

    2005

  59. [59]

    Gradient flows for sampling: Mean-field models, Gaussian approximations and affine invariance.arXiv preprint arXiv:2302.11024, 2023

    Yifan Chen, Daniel Zhengyu Huang, Jiaoyang Huang, Sebastian Reich, and Andrew M Stuart. Gradient flows for sampling: Mean-field models, Gaussian approximations and affine invariance.arXiv preprint arXiv:2302.11024, 2023

    work page arXiv 2023

  60. [60]

    Sampling via gradient flows in the space of probability measures.arXiv preprint arXiv:2310.03597, 2023

    Yifan Chen, Daniel Zhengyu Huang, Jiaoyang Huang, Sebastian Reich, and Andrew M Stuart. Sampling via gradient flows in the space of probability measures.arXiv preprint arXiv:2310.03597, 2023

    work page arXiv 2023

  61. [61]

    Kinetic equilibration rates for granular media and related equations: entropy dissipation and mass transportation estimates.Revista Matematica Iberoamericana, 19(3):971–1018, 2003

    José A Carrillo, Robert J McCann, and Cédric Villani. Kinetic equilibration rates for granular media and related equations: entropy dissipation and mass transportation estimates.Revista Matematica Iberoamericana, 19(3):971–1018, 2003

    2003

  62. [62]

    Aggregation-diffusion equations: dynamics, asymptotics, and singular limits

    José A Carrillo, Katy Craig, and Yao Yao. Aggregation-diffusion equations: dynamics, asymptotics, and singular limits. InActive Particles, Volume 2: Advances in Theory, Models, and Applications, pages 65–108. Springer, 2019

    2019

  63. [63]

    Propagation of chaos: a review of models, methods and applications

    Louis-Pierre Chaintron and Antoine Diez. Propagation of chaos: a review of models, methods and applications. I. models and methods.arXiv preprint arXiv:2203.00446, 2022

    work page arXiv 2022

  64. [64]

    Michele Ballerini, Nicola Cabibbo, Raphael Candelier, Andrea Cavagna, Evaristo Cisbani, Irene Giardina, Vivien Lecomte, Alberto Orlandi, Giorgio Parisi, Andrea Procaccini, et al. Interaction ruling animal collective behavior depends on topological rather than metric distance: Evidence from a field study.Proceedings of the national academy of sciences, 105...

    2008

  65. [65]

    Scale-free correlations in starling flocks.Proceedings of the National Academy of Sciences, 107(26):11865–11870, 2010

    Andrea Cavagna, Alessio Cimarelli, Irene Giardina, Giorgio Parisi, Raffaele Santagati, Fabio Stefanini, and Massimiliano Viale. Scale-free correlations in starling flocks.Proceedings of the National Academy of Sciences, 107(26):11865–11870, 2010

    2010

  66. [66]

    Collective motion and density fluctuations in bacterial colonies.Proceedings of the National Academy of Sciences, 107(31):13626–13630, 2010

    He-Peng Zhang, Avraham Be’er, E-L Florin, and Harry L Swinney. Collective motion and density fluctuations in bacterial colonies.Proceedings of the National Academy of Sciences, 107(31):13626–13630, 2010

    2010

  67. [67]

    TRex, a fast multi-animal tracking system with markerless identification, and 2d estimation of posture and visual fields.Elife, 10:e64000, 2021

    Tristan Walter and Iain D Couzin. TRex, a fast multi-animal tracking system with markerless identification, and 2d estimation of posture and visual fields.Elife, 10:e64000, 2021

    2021

  68. [68]

    ACCESS: Advancing innovation: NSF’s advanced cyberinfrastructure coordination ecosystem: Services & support

    Timothy J Boerner, Stephen Deems, Thomas R Furlani, Shelley L Knuth, and John Towns. ACCESS: Advancing innovation: NSF’s advanced cyberinfrastructure coordination ecosystem: Services & support. InPractice and experience in advanced research computing 2023: Computing for the common good, pages 173–176. Association for Computing Machinery, Inc, 2023

    2023

  69. [69]

    SIAM, 2016

    René Carmona.Lectures on BSDEs, stochastic control, and stochastic differential games with financial applica- tions. SIAM, 2016

    2016

  70. [70]

    V . S. Varadarajan. On the convergence of sample probability distributions.Sankhy ¯a: The Indian Journal of Statistics (1933-1960), 19(1/2):23–26, 1958. 23 Multiscale Nudge June 8, 2026

    1933

  71. [71]

    ∇ ˜Kh(¯Zi,N s − ¯Zk,N s )− ∇ ˜Kh ∗f t(¯Zi,N s ) ∇ ˜Kh(¯Zi,N s − ¯Zl,N s )− ∇ ˜Kh ∗f t(¯Zi,N s ) # ≤ λ2 N2 X k̸=l E

    C. Villani.Optimal Transport: Old and New. Grundlehren der mathematischen Wissenschaften. Springer Berlin Heidelberg, 2008. A Proof of Lemma 3.1 Proof. We first prove the regularity statements. The periodized Gaussian K T h is C ∞, periodic, even, and has unit mass on Td. Moreover, all of its derivatives are obtained by periodizing derivatives of the Eucl...

    2008