pith. sign in

arxiv: 1907.07810 · v1 · pith:WSTEH7WGnew · submitted 2019-07-17 · 🧮 math.NA · cs.LG· cs.NA· physics.data-an

Stability selection enables robust learning of partial differential equations from limited noisy data

Suryanarayana Maddu , Bevan L. Cheeseman , Ivo F. Sbalzarini , Christian L. M\"uller This is my paper

Pith reviewed 2026-05-24 19:54 UTC · model grok-4.3

classification 🧮 math.NA cs.LGcs.NAphysics.data-an
keywords PDE discoverysparse regressionstability selectionnoisy dataiterative hard-thresholdingmodel selectionreaction-diffusion
0
0 comments X

The pith

Stability selection with iterative hard-thresholding recovers partial differential equations from noisy data without manual tuning.

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 adds stability selection to sparse regression methods for identifying the governing partial differential equations in noisy spatiotemporal measurements. By using repeated subsampling to select the regularization strength that yields reproducible models, the approach removes the need for hand-tuned parameters while increasing tolerance to noise. The resulting method, when paired with iterative hard-thresholding, recovers accurate PDEs from smaller data sets than earlier techniques and is shown to work on both simulated benchmarks and real fluorescence data from an embryonic system.

Core claim

The combination of stability selection with the iterative hard-thresholding algorithm from compressed sensing provides a fast, parameter-free, and robust computational framework for PDE inference that outperforms previous algorithmic approaches with respect to recovery accuracy, amount of data required, and robustness to noise.

What carries the argument

PDE-STRIDE, a stability-based model selection procedure that determines the regularization level guaranteeing reproducible recovery of sparse support when combined with any penalized regression solver.

If this is right

  • Correct PDE terms are recovered from fewer spatial-temporal samples than required by earlier sparse regression schemes.
  • No user-specified regularization parameter is needed to obtain reproducible models.
  • The same pipeline recovers a reaction-diffusion-flow model directly from experimental microscopy images of C. elegans zygotes.
  • Model-component importance is ranked by selection frequency across subsamples, giving an interpretable stability score for each term.

Where Pith is reading between the lines

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

  • The approach could be tested on data sets where the library is deliberately over-complete to see how often extraneous terms survive the stability filter.
  • If the underlying dynamics are only approximately sparse, the method may still return the minimal stable model but the recovered coefficients would then represent an effective reduced description rather than the exact physics.
  • Extending the stability criterion to time-varying or parameter-dependent libraries might allow discovery of non-autonomous or spatially heterogeneous PDEs from the same data volumes.

Load-bearing premise

The true PDE is exactly sparse inside a pre-chosen library of candidate terms and repeated subsampling under the observed noise will identify that exact support without systematic bias from library choice or noise statistics.

What would settle it

Apply the method to simulated data generated from a known sparse PDE, add increasing levels of noise, and check whether the recovered support deviates from the true terms while the signal remains above the noise floor.

Figures

Figures reproduced from arXiv: 1907.07810 by Bevan L. Cheeseman, Christian L. M\"uller, Ivo F. Sbalzarini, Suryanarayana Maddu.

Figure 1
Figure 1. Figure 1: Enabling data-driven mathematical model discovery through stability selection: [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Model selection with PDE-STRIDE for the 1D Burgers equation: [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Comparison between different sparsity promoters for the 1D Burgers equation: Each colored square corresponds to a design (N, p, σ) with certain sample size N, disctionary size p, and nose level σ. Color indicates the success frequency over 30 independent repetitions with uniformly random data sub-samples. “Success” is defined as the existence of a λ for which the correct PDE is recovered from the data. The… view at source ↗
Figure 4
Figure 4. Figure 4: Model selection with PDE-STRIDE+IHT-d for 1D Burgers equation recovery : The top left image shows the numerical solution of the 1D Burgers equations on 256 × 100 space and time grid. The stability plots for the design N = 20, p = 19 show the separation of the true PDE components (in solid color) from the noisy components (dotted black). The inference power of the PDE-STRIDE method is tested for additive Ga… view at source ↗
Figure 5
Figure 5. Figure 5: Numerical solution of 2D Vorticity transport equation [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Model selection with PDE-STRIDE+IHT-d for 2D Vorticity transport equation recovery: The stability plots for the design N = 500, p = 48 show the separation of the true PDE components (in solid color) from the noisy components (dotted black). The inference power of the PDE-STRIDE method is tested for additive Gaussian noise-levels σ up-to 6% (not shown). In all the cases, perfect recovery was possible with t… view at source ↗
Figure 7
Figure 7. Figure 7: Model selection with PDE-STRIDE+IHT-d for 3D Gray-Scott u−component equation recovery : The top left figure shows the visualization of the 3D simulation domain with v species concentration. The color gradient corresponds to the varying concentration over space. The stability plots for the design N = 400, p = 69 show the separation of the true PDE components (solid color) from the noisy components (dotted b… view at source ↗
Figure 8
Figure 8. Figure 8: Achievability results for model selection with PDE-STRIDE [PITH_FULL_IMAGE:figures/full_fig_p014_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Achievability results for model selection with PDE-STRIDE [PITH_FULL_IMAGE:figures/full_fig_p015_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Data-driven model inference of the regulatory network of membrane PAR proteins from spatiotemporal data [PITH_FULL_IMAGE:figures/full_fig_p016_10.png] view at source ↗
read the original abstract

We present a statistical learning framework for robust identification of partial differential equations from noisy spatiotemporal data. Extending previous sparse regression approaches for inferring PDE models from simulated data, we address key issues that have thus far limited the application of these methods to noisy experimental data, namely their robustness against noise and the need for manual parameter tuning. We address both points by proposing a stability-based model selection scheme to determine the level of regularization required for reproducible recovery of the underlying PDE. This avoids manual parameter tuning and provides a principled way to improve the method's robustness against noise in the data. Our stability selection approach, termed PDE-STRIDE, can be combined with any sparsity-promoting penalized regression model and provides an interpretable criterion for model component importance. We show that in particular the combination of stability selection with the iterative hard-thresholding algorithm from compressed sensing provides a fast, parameter-free, and robust computational framework for PDE inference that outperforms previous algorithmic approaches with respect to recovery accuracy, amount of data required, and robustness to noise. We illustrate the performance of our approach on a wide range of noise-corrupted simulated benchmark problems, including 1D Burgers, 2D vorticity-transport, and 3D reaction-diffusion problems. We demonstrate the practical applicability of our method on real-world data by considering a purely data-driven re-evaluation of the advective triggering hypothesis for an embryonic polarization system in C.~elegans. Using fluorescence microscopy images of C.~elegans zygotes as input data, our framework is able to recover the PDE model for the regulatory reaction-diffusion-flow network of the associated proteins.

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 manuscript introduces PDE-STRIDE, a stability selection framework that augments sparse regression methods (in particular iterative hard-thresholding) to identify governing PDEs from noisy spatiotemporal data. It claims to eliminate manual regularization tuning, improve robustness to noise, reduce data requirements, and outperform prior approaches on 1D Burgers, 2D vorticity-transport, and 3D reaction-diffusion benchmarks as well as on fluorescence data from a C. elegans embryonic polarization system.

Significance. If the central claims hold, the work supplies a practical, interpretable tool for data-driven PDE discovery that directly addresses two long-standing obstacles—noise sensitivity and hyperparameter selection—in sparse regression pipelines. The stability scores provide an explicit importance measure for candidate terms, and the combination with IHT yields a computationally efficient procedure. The real-data example demonstrates applicability beyond synthetic test cases.

major comments (2)
  1. [§4] §4 (Benchmark problems): All reported numerical experiments generate data from the exact sparse combination of terms that is later placed in the candidate library. Consequently the recovery-accuracy and noise-robustness gains are demonstrated only under perfect library specification; the manuscript contains no experiments that introduce small omitted terms or an incomplete library, which directly limits the strength of the robustness claim.
  2. [§3.2] §3.2 (Stability selection procedure): The method presupposes that the true PDE is exactly sparse inside the pre-specified library. While stability selection mitigates noise-induced variability in support recovery, the paper provides neither theoretical bounds nor numerical tests that quantify how the stability scores degrade when this exact-sparsity assumption is mildly violated (e.g., by a small unmodeled forcing term).
minor comments (2)
  1. [§4] Figure captions in §4 would benefit from explicit statements of the noise-to-signal ratios and the number of spatial-temporal points used in each trial.
  2. [Abstract] The abstract states that the method “outperforms previous algorithmic approaches” but does not name the baselines or report quantitative margins; a short table summarizing these comparisons would improve clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive comments. We address each major point below and will revise the manuscript to incorporate the suggested clarifications and additional experiments.

read point-by-point responses

  1. Referee: [§4] §4 (Benchmark problems): All reported numerical experiments generate data from the exact sparse combination of terms that is later placed in the candidate library. Consequently the recovery-accuracy and noise-robustness gains are demonstrated only under perfect library specification; the manuscript contains no experiments that introduce small omitted terms or an incomplete library, which directly limits the strength of the robustness claim.

    Authors: We agree that the reported benchmarks assume the true PDE lies exactly in the span of a sparse subset of the candidate library. This is the standard setting for sparse-regression PDE discovery, and our stability-selection procedure is intended to improve noise robustness within that setting. Model misspecification arising from an incomplete library is a separate and important issue. To address the referee’s concern and strengthen the robustness claims, we will add a new set of numerical experiments that introduce small omitted terms (e.g., a weak unmodeled forcing) and quantify the resulting degradation in stability scores and recovery accuracy. revision: yes

  2. Referee: [§3.2] §3.2 (Stability selection procedure): The method presupposes that the true PDE is exactly sparse inside the pre-specified library. While stability selection mitigates noise-induced variability in support recovery, the paper provides neither theoretical bounds nor numerical tests that quantify how the stability scores degrade when this exact-sparsity assumption is mildly violated (e.g., by a small unmodeled forcing term).

    Authors: The framework is formulated under the exact-sparsity assumption that is standard for sparse regression methods; stability selection is shown to reduce variability due to noise under this assumption. We do not derive theoretical bounds on performance under mild violations of exact sparsity, as such analysis lies outside the scope of the present work. We will, however, add numerical experiments that introduce small unmodeled terms and report the empirical effect on stability scores, thereby providing the quantitative assessment requested by the referee. revision: yes

Circularity Check

0 steps flagged

No circularity in the algorithmic framework or derivation chain.

full rationale

The paper presents PDE-STRIDE as a new stability-selection wrapper around existing sparsity-promoting regressors (including IHT from compressed sensing) to select regularization level without manual tuning. Performance claims rest on empirical benchmarks (Burgers, vorticity, reaction-diffusion) and one real-world dataset, not on any derivation that reduces a result to its own fitted inputs or self-citations. The explicit modeling assumption of exact sparsity in a pre-specified library is stated up-front rather than smuggled in via self-reference or tautological redefinition. No load-bearing step equates a prediction to a fit by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The approach inherits the standard sparsity assumption of dictionary-based regression for PDEs and relies on the empirical effectiveness of stability selection under the noise present in the target data; no new entities are postulated.

axioms (1)
  • domain assumption The governing PDE can be expressed as a sparse linear combination of terms drawn from a pre-defined candidate library.
    This is the foundational premise of all sparse regression methods for PDE identification mentioned in the abstract.

pith-pipeline@v0.9.0 · 5845 in / 1230 out tokens · 20978 ms · 2026-05-24T19:54:52.895667+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

74 extracted references · 74 canonical work pages · 1 internal anchor

  1. [1]

    Quantitative modeling in cell biology: what is it good for? Developmental cell, 11(3):279–287, 2006

    Alex Mogilner, Roy Wollman, and Wallace F Marshall. Quantitative modeling in cell biology: what is it good for? Developmental cell, 11(3):279–287, 2006

    work page 2006

  2. [2]

    Modeling and simulation of biological systems from image data

    Ivo F Sbalzarini. Modeling and simulation of biological systems from image data. Bioessays, 35(5): 482–490, 2013

    work page 2013

  3. [3]

    Biology by numbers: mathematical modelling in developmental biology

    Claire J Tomlin and Je ffrey D Axelrod. Biology by numbers: mathematical modelling in developmental biology. Nature reviews genetics, 8(5):331, 2007

    work page 2007

  4. [4]

    Finite Difference methods in financial engineering: a Partial Di fferential Equation ap- proach

    Daniel J Du ffy. Finite Difference methods in financial engineering: a Partial Di fferential Equation ap- proach. John Wiley & Sons, 2013

    work page 2013

  5. [5]

    Solving the mathematical models of neurosciences and medicine

    George Adomian. Solving the mathematical models of neurosciences and medicine. Mathematics and computers in simulation, 40(1-2):107–114, 1995

    work page 1995

  6. [6]

    50 years of data science

    David Donoho. 50 years of data science. Journal of Computational and Graphical Statistics, 26(4):745– 766, 2017

    work page 2017

  7. [7]

    Bayesian design of synthetic biological systems

    Chris P Barnes, Daniel Silk, Xia Sheng, and Michael P H Stumpf. Bayesian design of synthetic biological systems. Proceedings of the National Academy of Sciences of the United States of America , 108(37): 15190–15195, 2011. ISSN 0027-8424. doi: 10.1073 /pnas.1017972108

    work page 2011

  8. [8]

    L p-Adaptation: Simultaneous Design Centering and Robustness Estimation of Electronic and Biological Systems

    Josefine Asmus, Christian L M¨uller, and Ivo F Sbalzarini. L p-Adaptation: Simultaneous Design Centering and Robustness Estimation of Electronic and Biological Systems. Scientific Reports , 7(1):6660, 2017. ISSN 20452322. doi: 10.1038 /s41598-017-03556-5

    work page 2017

  9. [9]

    Computational systems biology

    Hiroaki Kitano. Computational systems biology. Nature, 420(6912):206, 2002

    work page 2002

  10. [10]

    Diffusion and scaling during early embryonic pattern formation

    Thomas Gregor, William Bialek, Rob R de Ruyter van Steveninck, David W Tank, and Eric F Wieschaus. Diffusion and scaling during early embryonic pattern formation. Proceedings of the National Academy of Sciences, 102(51):18403–18407, 2005

    work page 2005

  11. [11]

    Modeling gene expression with di fferential equations

    Ting Chen, Hongyu L He, and George M Church. Modeling gene expression with di fferential equations. In Biocomputing’99, pages 29–40. World Scientific, 1999

    work page 1999

  12. [12]

    Mathematical modeling of gene expression: a guide for the perplexed biologist

    Ahmet Ay and David N Arnosti. Mathematical modeling of gene expression: a guide for the perplexed biologist. Critical reviews in biochemistry and molecular biology, 46(2):137–151, 2011

    work page 2011

  13. [13]

    Active gel physics

    Jacques Prost, Frank J ¨ulicher, and Jean-Franc ¸ois Joanny. Active gel physics. Nature Physics, 11(2):111, 2015

    work page 2015

  14. [14]

    Attachment of the blastoderm to the vitelline envelope a ffects gastrulation of insects

    Stefan M ¨unster, Akanksha Jain, Alexander Mietke, Anastasios Pavlopoulos, Stephan W Grill, and Pavel Tomancak. Attachment of the blastoderm to the vitelline envelope a ffects gastrulation of insects. Nature, page 1, 2019

    work page 2019

  15. [15]

    Identification of continuous, spatiotemporal systems

    H V oss, MJ B ¨unner, and Markus Abel. Identification of continuous, spatiotemporal systems. Physical Review E, 57(3):2820, 1998

    work page 1998

  16. [16]

    Friedman

    Leo Breiman and Jerome H. Friedman. Estimating optimal transformations for multiple regression and cor- relation. Journal of the American Statistical Association, 1985. ISSN 1537274X. doi: 10.1080 /01621459. 1985.10478157

    work page arXiv 1985

  17. [17]

    Fitting partial differential equations to space-time dynam- ics

    Markus B ¨ar, Rainer Hegger, and Holger Kantz. Fitting partial differential equations to space-time dynam- ics. Physical Review E, 59(1):337, 1999

    work page 1999

  18. [18]

    Xiaolei Xun, Jiguo Cao, Bani Mallick, Arnab Maity, and Raymond J. Carroll. Parameter estimation of partial differential equation models. Journal of the American Statistical Association, 108(503):1009–1020,

    work page

  19. [19]

    doi: 10.1080 /01621459.2013.794730

    ISSN 01621459. doi: 10.1080 /01621459.2013.794730. 18

    work page arXiv 2013

  20. [20]

    Machine learning of linear di fferential equa- tions using gaussian processes

    Maziar Raissi, Paris Perdikaris, and George Em Karniadakis. Machine learning of linear di fferential equa- tions using gaussian processes. Journal of Computational Physics, 348:683–693, 2017

    work page 2017

  21. [21]

    Discovering governing equations from data by sparse identification of nonlinear dynamical systems

    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 , page 201517384, 2016

    work page 2016

  22. [22]

    Data-driven discovery of partial differential equations

    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

    work page 2017

  23. [23]

    Learning partial differential equations via data discovery and sparse optimization

    Hayden Schae ffer. Learning partial differential equations via data discovery and sparse optimization. Proc. R. Soc. A, 473(2197):20160446, 2017

    work page 2017

  24. [24]

    Robust data-driven discovery of governing physical laws with error bars.Pro- ceedings of the Royal Society A: Mathematical, Physical and Engineering Science , 474(2217):20180305,

    Sheng Zhang and Guang Lin. Robust data-driven discovery of governing physical laws with error bars.Pro- ceedings of the Royal Society A: Mathematical, Physical and Engineering Science , 474(2217):20180305,

    work page

  25. [25]

    doi: 10.1098/rspa.2018.0305

    ISSN 1364-5021. doi: 10.1098/rspa.2018.0305. URL http://rspa.royalsocietypublishing. org/lookup/doi/10.1098/rspa.2018.0305

    work page doi:10.1098/rspa.2018.0305 2018

  26. [26]

    Model selection for hybrid dynamical systems via sparse regression

    N M Mangan, T Askham, S L Brunton, J N Kutz, and J L Proctor. Model selection for hybrid dynamical systems via sparse regression. Proceedings of the Royal Society A: Mathematical, Physical and Engineer- ing Sciences, 475(2223):1–22, 2019. ISSN 14712946. doi: 10.1098 /rspa.2018.0534

    work page arXiv 2019

  27. [27]

    PDE-Net: Learning PDEs from Data

    Zichao Long, Yiping Lu, Xianzhong Ma, and Bin Dong. PDE-Net: Learning PDEs from data. arXiv preprint arXiv:1710.09668, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  28. [28]

    Hidden physics models: Machine learning of nonlinear partial differential equations

    Maziar Raissi and George Em Karniadakis. Hidden physics models: Machine learning of nonlinear partial differential equations. Journal of Computational Physics, 357:125–141, 2018

    work page 2018

  29. [29]

    Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial dif- ferential equations

    M Raissi, P Perdikaris, and G E Karniadakis. Physics-informed neural networks: A deep learning frame- work for solving forward and inverse problems involving nonlinear partial di fferential equations. Journal of Computational Physics , 378:686–707, 2019. ISSN 10902716. doi: 10.1016 /j.jcp.2018.10.045. URL https://doi.org/10.1016/j.jcp.2018.10.045

    work page doi:10.1016/j.jcp.2018.10.045 2019

  30. [30]

    PDE-Net 2.0: Learning PDEs from data with a numeric-symbolic hybrid deep network

    Zichao Long, Yiping Lu, and Bin Dong. PDE-Net 2.0: Learning PDEs from data with a numeric-symbolic hybrid deep network. arXiv preprint arXiv:1812.04426, 2018

    work page arXiv 2018

  31. [31]

    Image restoration: Wavelet frame shrinkage, nonlinear evolution pdes, and beyond

    Bin Dong, Qingtang Jiang, and Zuowei Shen. Image restoration: Wavelet frame shrinkage, nonlinear evolution pdes, and beyond. Multiscale Modeling & Simulation, 15(1):606–660, 2017

    work page 2017

  32. [32]

    Stability selection

    Nicolai Meinshausen and Peter B ¨uhlmann. Stability selection. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 72(4):417–473, 2010

    work page 2010

  33. [33]

    Variable selection with error control: another look at stability selection

    Rajen D Shah and Richard J Samworth. Variable selection with error control: another look at stability selection. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 75(1):55–80, 2013

    work page 2013

  34. [34]

    Regression shrinkage and selection via the lasso

    Robert Tibshirani. Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society. Series B (Methodological), pages 267–288, 1996

    work page 1996

  35. [35]

    Iterative thresholding for sparse approximations

    Thomas Blumensath and Mike E Davies. Iterative thresholding for sparse approximations. Journal of Fourier analysis and Applications, 14(5-6):629–654, 2008

    work page 2008

  36. [36]

    Hard thresholding pursuit: an algorithm for compressive sensing

    Simon Foucart. Hard thresholding pursuit: an algorithm for compressive sensing. SIAM Journal on Nu- merical Analysis, 49(6):2543–2563, 2011

    work page 2011

  37. [37]

    Guiding self-organized pattern formation in cell polarity establishment

    Peter Gross, K Vijay Kumar, Nathan W Goehring, Justin S Bois, Carsten Hoege, Frank J ¨ulicher, and Stephan W Grill. Guiding self-organized pattern formation in cell polarity establishment. Nature Physics, 15(3):293, 2019

    work page 2019

  38. [38]

    Numerical di fferentiation of noisy, nonsmooth data

    Rick Chartrand. Numerical di fferentiation of noisy, nonsmooth data. ISRN Applied Mathematics , 2011, 2011. 19

    work page 2011

  39. [39]

    Data smoothing and numerical di fferentiation by a regularization method

    Jonathan J Stickel. Data smoothing and numerical di fferentiation by a regularization method. Computers & chemical engineering, 34(4):467–475, 2010

    work page 2010

  40. [40]

    Coordinate descent algorithms for lasso penalized regression

    Tong Tong Wu and Kenneth Lange. Coordinate descent algorithms for lasso penalized regression. Annals of Applied Statistics, 2(1):224–244, 2008. ISSN 19326157. doi: 10.1214 /07-AOAS147

    work page 2008

  41. [41]

    Regularization paths for generalized linear models via coordinate descent

    Jerome Friedman, Trevor Hastie, and Rob Tibshirani. Regularization paths for generalized linear models via coordinate descent. Journal of statistical software, 33(1):1, 2010

    work page 2010

  42. [42]

    On the douglasrachford splitting method and the proximal point algorithm for maximal monotone operators

    Jonathan Eckstein and Dimitri P Bertsekas. On the douglasrachford splitting method and the proximal point algorithm for maximal monotone operators. Mathematical Programming, 55(1-3):293–318, 1992

    work page 1992

  43. [43]

    Proximal splitting methods in signal processing

    Patrick L Combettes and Jean-Christophe Pesquet. Proximal splitting methods in signal processing. In Fixed-point algorithms for inverse problems in science and engineering, pages 185–212. Springer, 2011

    work page 2011

  44. [44]

    A Fast Iterative Shrinkage-Thresholding Algorithm

    Amir Beck and Marc Teboulle. A Fast Iterative Shrinkage-Thresholding Algorithm. Society for Industrial and Applied Mathematics Journal on Imaging Sciences , 2(1):183–202, 2009. ISSN 1936-4954. doi: 10. 1137/080716542

    work page 2009

  45. [45]

    High-dimensional graphs and variable selection with the lasso

    Nicolai Meinshausen, Peter B ¨uhlmann, et al. High-dimensional graphs and variable selection with the lasso. The annals of statistics, 34(3):1436–1462, 2006

    work page 2006

  46. [46]

    On model selection consistency of lasso

    Peng Zhao and Bin Yu. On model selection consistency of lasso. Journal of Machine learning research, 7 (Nov):2541–2563, 2006

    work page 2006

  47. [47]

    Variable Selection via Nonconcave Penalized Likelihood and its Oracle Prop- erties

    Jianqing Fan and Runze Li. Variable Selection via Nonconcave Penalized Likelihood and its Oracle Prop- erties. Journal of the American Statistical Association, 96(456):1348–1360, 2001. ISSN 0162-1459. doi: 10.1198/016214501753382273

    work page doi:10.1198/016214501753382273 2001

  48. [48]

    Nearly unbiased variable selection under minimax concave penalty , volume 38

    Cun Hui Zhang. Nearly unbiased variable selection under minimax concave penalty , volume 38. 2010. ISBN 9040210063. doi: 10.1214 /09-AOS729

    work page 2010

  49. [49]

    Greed is good: Algorithmic results for sparse approximation

    Joel A Tropp. Greed is good: Algorithmic results for sparse approximation. IEEE Transactions on Infor- mation theory, 50(10):2231–2242, 2004

    work page 2004

  50. [50]

    CoSaMP: Iterative signal recovery from incomplete and inaccurate samples

    Deanna Needell and Joel A Tropp. CoSaMP: Iterative signal recovery from incomplete and inaccurate samples. Applied and computational harmonic analysis, 26(3):301–321, 2009

    work page 2009

  51. [51]

    Subspace pursuit for compressive sensing signal reconstruction

    Wei Dai and Olgica Milenkovic. Subspace pursuit for compressive sensing signal reconstruction. IEEE transactions on Information Theory, 55(5):2230–2249, 2009

    work page 2009

  52. [52]

    Iterative hard thresholding for compressed sensing

    Thomas Blumensath and Mike E Davies. Iterative hard thresholding for compressed sensing. Applied and computational harmonic analysis, 27(3):265–274, 2009

    work page 2009

  53. [53]

    Sparse approximation via iterative thresholding

    Kyle K Herrity, Anna C Gilbert, and Joel A Tropp. Sparse approximation via iterative thresholding. In 2006 IEEE International Conference on Acoustics Speech and Signal Processing Proceedings , volume 3, pages III–III. IEEE, 2006

    work page 2006

  54. [54]

    Just relax: Convex programming methods for identifying sparse signals in noise

    Joel A Tropp. Just relax: Convex programming methods for identifying sparse signals in noise. IEEE transactions on information theory, 52(3):1030–1051, 2006

    work page 2006

  55. [55]

    Gradient projection for sparse reconstruc- tion: Application to compressed sensing and other inverse problems

    M ´ario AT Figueiredo, Robert D Nowak, and Stephen J Wright. Gradient projection for sparse reconstruc- tion: Application to compressed sensing and other inverse problems. IEEE Journal of selected topics in signal processing, 1(4):586–597, 2007

    work page 2007

  56. [56]

    A Study of Cross-Validation and Bootstrap for Accuracy Estimation and Model Selection

    Ron Kohavi. A Study of Cross-Validation and Bootstrap for Accuracy Estimation and Model Selection. International Joint Conference of Artificial Intelligence, 1995

    work page 1995

  57. [57]

    The Annals of Statistics , author =

    Gideon Schwarz. Estimating the Dimension of a Model. The Annals of Statistics, 1978. ISSN 0090-5364. doi: 10.1214/aos/1176344136. 20

    work page doi:10.1214/aos/1176344136 1978

  58. [58]

    M ¨uller

    Johannes Lederer and Christian L. M ¨uller. Don’t fall for tuning parameters: Tuning-free variable selection in high dimensions with the TREX. In Proceedings of the Twenty-Ninth AAAI Conference on Artificial Intelligence (AAAI 2015), pages 2729—-2735. AAAI Press, 2015

    work page 2015

  59. [59]

    M ¨uller

    Jacob Bien, Irina Gaynanova, Johannes Lederer, and Christian L. M ¨uller. Non-Convex Global Min- imization and False Discovery Rate Control for the TREX. Journal of Computational and Graph- ical Statistics , 27(1):23–33, 2018. ISSN 15372715. doi: 10.1080 /10618600.2017.1341414. URL http://arxiv.org/abs/1604.06815

    work page arXiv 2018

  60. [60]

    Stability

    Bin Yu. Stability. Bernoulli, 19(4):1484–1500, 2013. ISSN 1350-7265. doi: 10.3150 /13-BEJSP14. URL http://projecteuclid.org/euclid.bj/1377612862

    work page arXiv 2013

  61. [61]

    Stability Approach to Regularization Selection (StARS) for High Dimensional Graphical Models

    Han Liu, Kathryn Roeder, and Larry Wasserman. Stability Approach to Regularization Selection (StARS) for High Dimensional Graphical Models. Advances in neural information processing systems, 24(2):1432–1440, 2010. ISSN 1049-5258. URL https://papers.nips.cc/paper/ 3966-stability-approach-to-regularization-selection-stars-for-high-dimensional-graphical-mode...

    work page arXiv 2010

  62. [62]

    High-dimensional statistics with a view toward appli- cations in biology

    Peter B ¨uhlmann, Markus Kalisch, and Lukas Meier. High-dimensional statistics with a view toward appli- cations in biology. Annual Review of Statistics and Its Application, 1:255–278, 2014

    work page 2014

  63. [63]

    Random lasso

    Sijian Wang, Bin Nan, Saharon Rosset, and Ji Zhu. Random lasso. The annals of applied statistics, 5(1): 468, 2011

    work page 2011

  64. [64]

    Navier-Stokes equations: theory and numerical analysis , volume 343

    Roger Temam. Navier-Stokes equations: theory and numerical analysis , volume 343. American Mathe- matical Soc., 2001

    work page 2001

  65. [65]

    OpenFPM: A scalable open framework for particle and particle-mesh codes on parallel computers

    Pietro Incardona, Antonio Leo, Yaroslav Zaluzhnyi, Rajesh Ramaswamy, and Ivo F Sbalzarini. OpenFPM: A scalable open framework for particle and particle-mesh codes on parallel computers. Computer Physics Communications, 2019

    work page 2019

  66. [66]

    Models of biological pattern formation

    Hans Meinhardt. Models of biological pattern formation. New York, 1982

    work page 1982

  67. [67]

    A living mesoscopic cellular automaton made of skin scales

    Liana Manukyan, Sophie A Montandon, Anamarija Fofonjka, Stanislav Smirnov, and Michel C Milinkovitch. A living mesoscopic cellular automaton made of skin scales. Nature, 544(7649):173, 2017

    work page 2017

  68. [68]

    Sharp thresholds for high-dimensional and noisy sparsity recovery using l1- con- strained quadratic programming (lasso)

    Martin J Wainwright. Sharp thresholds for high-dimensional and noisy sparsity recovery using l1- con- strained quadratic programming (lasso). IEEE transactions on information theory, 55(5):2183–2202, 2009

    work page 2009

  69. [69]

    Asymmetrically distributed PAR-3 protein contributes to cell polarity and spindle alignment in early C

    Bijan Etemad-Moghadam, Su Guo, and Kenneth J Kemphues. Asymmetrically distributed PAR-3 protein contributes to cell polarity and spindle alignment in early C. elegans embryos. Cell, 83(5):743–752, 1995

    work page 1995

  70. [70]

    Polarization of PAR proteins by advective triggering of a pattern- forming system

    Nathan W Goehring, Philipp Khuc Trong, Justin S Bois, Debanjan Chowdhury, Ernesto M Nicola, An- thony A Hyman, and Stephan W Grill. Polarization of PAR proteins by advective triggering of a pattern- forming system. Science, 334(6059):1137–1141, 2011

    work page 2011

  71. [71]

    Hierarchical grouping to optimize an objective function

    Joe H Ward Jr. Hierarchical grouping to optimize an objective function. Journal of the American statistical association, 58(301):236–244, 1963

    work page 1963

  72. [72]

    Data-driven discovery of gov- erning physical laws and their parametric dependencies in engineering, physics and biology

    J Nathan Kutz, Samuel H Rudy, Alessandro Alla, and Steven L Brunton. Data-driven discovery of gov- erning physical laws and their parametric dependencies in engineering, physics and biology. In2017 IEEE 7th International Workshop on Computational Advances in Multi-Sensor Adaptive Processing (CAMSAP), pages 1–5. IEEE, 2017

    work page 2017

  73. [73]

    The optimal hard threshold for singular values is 4 / 3, 2013

    David L Donoho and Matan Gavish. The optimal hard threshold for singular values is 4 / 3, 2013

    work page 2013

  74. [74]

    Truncated singular value decomposition solutions to discrete ill-posed problems with ill-determined numerical rank

    Per Christian Hansen. Truncated singular value decomposition solutions to discrete ill-posed problems with ill-determined numerical rank. SIAM Journal on Scientific and Statistical Computing, 11(3):503–518, 1990. 21 6 Supplementary Material 6.1 Algorithm Algorithm 1 ˆξ = arg minξ∥Ut− Θξ∥2 2 +λ∥ξ∥0 Problem: ˆξ = arg minξ∥Ut− Θξ∥2 2 +λ∥ξ∥0 IHD-d(Θ, Ut,λ, max...

    work page 1990