Pith. sign in

REVIEW 2 major objections 4 minor 66 references

Linear scaling causal discovery from high-dimensional time series by dynamical community detection

T0 review · 2 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read This paper claims a causal-discovery framework whose computational cost scales linearly with the number of time-series variables, by optimizing the Information Imbalance to find dynamical communities and then ordering them into a…

desk verdict The community detection extension is real, but the linear-scaling claim is contradicted by the algorithm as written. read the letter →

arxiv 2501.10886 v2 pith:36D5XPAX submitted 2025-01-18 physics.data-an stat.ME

Matteo Allione , Vittorio Del Tatto , Alessandro Laio This is my paper · ORCID
classification physics.data-anstat.ME
keywords causaldiscoverytimeseriesInformationImbalancedynamicalcommunitiescommunitygraphlinearscalingdifferentiablehigh-dimensionaldynamics
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper argues that causal discovery in high-dimensional time series can be done at a computational cost that grows only linearly with the number of variables. The key step is to infer, for each variable, the set of variables that influence it by minimizing the differentiable Information Imbalance over one weight per variable, then to group variables that influence each other into 'dynamical communities' and order these communities by autonomy. The output is a coarse-grained community causal graph. On three synthetic systems (15 coupled logistic maps, five coupled Lorenz oscillators, and two coupled Lorenz-96 systems totaling 80 variables) the recovered graph matches the ground truth over a wide range of the threshold parameter, and the cost stays linear because the method performs one optimization per variable instead of an exponential search over conditioning sets.

What carries the argument

The machinery is the Differentiable Information Imbalance (DII), $$\mathrm{DII}(d_A\to d_B)=\frac{2}{$N^{2}$}\sum_{i\neq j}\frac{$e^{{-d_{ij}}$^A/\$\lambda$}}{\sum_{m\neq i}$e^{{-d_{im}}$^A/\$\lambda$}}r_{ij}^B,$$ a soft version of the rank-based Information Imbalance, recovered as $\lambda\to0$. Because it is differentiable in the distances, the weights $w$ entering $d_A$ can be optimized by gradient descent. The algorithm computes, for each target variable $X^\beta$, the optimal weights in a distance built from all variables at time $0$ predicting $X^\beta(\tau)$, takes the maximum over lags $\tau$ up to the autocorrelation time, thresholds the resulting matrix $G_{\alpha\beta}$ at $\varepsilon$, and closes under ancestor relations. Iteratively extracting minimal autonomous sets gives the dynamical communities and their autonomy levels; directed arrows between consecutive levels form the final community causal graph.

What would settle it

An experiment that would settle whether the causal criterion holds: generate data where a true direct cause's influence is exactly canceled by an opposing path, a violation of faithfulness, and check whether its optimized weight is zero. If the weight is zero while the variable is genuinely causal, the criterion of Eq. (2) fails and the method cannot recover the graph; if it stays nonzero, the criterion survives a harsher test than the paper's noise-additive experiments.

Watch

Extended reading notes

Core claim

On the paper's own terms, the claim is that optimizing the differentiable Information Imbalance against each target variable, once per variable, produces a connectivity matrix $G$ whose thresholded and ancestor-closed pattern reveals the system's dynamical communities. A community is a minimal autonomous set: a group of variables whose influence sets coincide, so that each member depends on all and only the other members, and the group's evolution is independent of the rest of the network. Removing these communities iteratively yields autonomy levels, and directed arrows between communities at consecutive levels form a directed acyclic community causal graph. The paper reports that this procedure recovers the exact community graph for its three test systems over a wide range of thresholds, and that a separate conditional test can decide whether a link between non-consecutive communities is direct or indirect.

Load-bearing premise

The load-bearing premise is the imported, unproved rule that a variable gets a nonzero optimal weight exactly when it causes the target; if hidden common drivers, canceling influences, or a bad local minimum break that rule, the communities and causal graph built on it are unreliable.

Editorial extensions

If this is right

  • If the claim is right, causal discovery on systems with hundreds of variables needs only one gradient-descent optimization per variable, avoiding the exponential search over conditioning sets.
  • The recovered graph is mesoscopic: variables that mutually influence each other are merged into a single node, so the output is a coarse-grained summary of the system's causal organization.
  • On the paper's three test systems the correct community graph is recovered for a wide range of the threshold $\varepsilon$, so the result is not tied to a finely tuned hyperparameter value.
  • Links between communities at non-consecutive autonomy levels can be classified as direct or indirect by an additional conditional test, refining the community graph.
  • The connectivity matrix $G$ can seed or speed up other causal-discovery and graph-clustering approaches, including cases with bidirectional couplings.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The 'linear scaling' is in the number of variables; each DII evaluation still sums over pairs of sampled frames, so total cost also grows with trajectory length and sample size, which matters for very long recordings.
  • The causal criterion of Eq. (2) is the unproved load-bearing step; the synthetic tests add noise to avoid deterministic non-faithfulness, so they do not probe whether the criterion fails under cancellations or with real confounded data.
  • A natural extension the paper only hints at is to use the inferred communities as a prior to constrain the conditioning sets of constraint-based algorithms, potentially combining the linear scaling with stronger statistical guarantees.
  • In observational data with unobserved common drivers, the paper's own supplemental analysis predicts that distinct true communities will be merged; that prediction could be tested directly on datasets with known confounders.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 4 minor

Summary. This paper introduces a causal-discovery framework for high-dimensional time series that groups variables into 'dynamical communities' by optimizing a differentiable Information Imbalance (DII). For each target variable β, Eq. (5) minimizes DII over weights assigned to all variables at t = 0; the resulting matrix G (Eq. 6) is thresholded to obtain autonomous sets, which are then organized hierarchically into communities and a community causal graph. The method is validated on coupled logistic maps, five coupled Lorenz systems, and two coupled Lorenz-96 systems, with exact reconstruction over wide threshold ranges. The paper claims that the computational cost scales linearly with the number of variables D.

Significance. If the efficiency claim were supported, the community-level coarse-graining would be a useful and novel tool for high-dimensional causal discovery. The paper's strengths are its clean synthetic validation with external ground truth, the demonstration of exact reconstruction over wide threshold ranges, the comparison with PCMCI, and the supplementary analysis of observational noise and causal sufficiency. However, the central claim of linear scaling is not supported by the algorithm as written, and the causal criterion is imported from prior work without a proof. Both points must be resolved before publication.

major comments (2)
  1. [§i and Discussion, Eq. (5)] The paper's headline claim that the computational cost scales linearly with D is not what the algorithm delivers. Eq. (5) performs one DII minimization per target variable β, but each minimization optimizes a weight vector of length D (Eq. 3), so computing the DII of Eq. (4) requires, per gradient epoch, pairwise distances over all D variables: with N = 2000 and mini-batches of 100 points, each batch is O(100^2 D) and 20 batches per epoch are O(200,000 D). Across D targets and E epochs the total is Θ(E N^2 D^2) (up to mini-batch constants), not O(D). Appendix A adds a further quadratic-in-communities term. The Discussion only claims that the number of optimizations scales linearly, which is true but does not imply linear total cost. The authors should either provide a rigorous complexity analysis with a sparse or block implementation, or revise the abstract and introduction so that the claim is limited to the number of optimizations.
  2. [SM §I, Eq. (S6)] The criterion that Xα causes Xβ if and only if the optimal weight component w^α_β is nonzero is assumed, not derived, and it is inherited from ref. [19] as a 'generalization'. The paper does not prove conditions under which this holds (faithfulness, absence of local minima in the DII optimization, causal sufficiency), and the synthetic tests are not designed to stress-test it. I recommend stating explicitly that this is an assumption, citing the prior analysis, and discussing failure modes; otherwise the community graph inherits any error in this step.
minor comments (4)
  1. [Title] The title contains a spacing typo: 'dynami cal' should be 'dynamical'.
  2. [Appendix B, §3] The sentence 'Eqs. (12) were integraterd with time step dt = 0.03' contains a typo: 'integraterd' should be 'integrated'.
  3. [§iii and Fig. 2] In the caption of Fig. 2, 'recollected ones' should be 'recovered ones' for clarity.
  4. [§ii] The definition of a minimal autonomous set, 'for every variable xα ∈ Sβ, one has Sα ≡ Sβ', is clear for exact sets but the thresholding step that produces the Sβ from G is not described with the same precision; a short explanation of how ties and near-equal sets are handled would help reproducibility.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: causal criterion is an explicit same-group assumption with external benchmark support, and community detection is validated against independent ground truth; the O(D) cost mismatch is a correctness concern, not circularity.

full rationale

The closest candidate for circularity is the causality criterion of Eq. (2), reused from the same group's ref. [19] and then generalized in Eq. (5) as 'X alpha is a direct or indirect cause of X beta when the alpha-component of w-beta is nonzero.' The paper does not present this as a theorem derived here; it explicitly says 'we assumed that if X causes Y ...' and the Supplemental Material frames the DII test as a conditional-independence test. The reuse is load-bearing, but it is an external, published, falsifiable result, and the present paper validates the entire pipeline on synthetic systems with independently known ground-truth connectivity, so the demonstration does not reduce to the citation. The community construction in step ii is a deterministic aggregation of the matrix G, not a fit to ground truth; the threshold epsilon is either scanned in validation plots showing wide plateaus of exact reconstruction or set to a data-dependent average, so the conclusion is not forced by tuning epsilon to the labels. The paper's advertised 'linear scaling' is not supported by the algorithm as written, since Eq. (5) runs D DII minimizations each over D weights, giving a quadratic total cost in D, and Appendix A can add a further quadratic term in the number of communities; however, that is a complexity/soundness issue rather than a circular reduction. No equation in the paper is equivalent to its input by construction, and no fitted parameter is renamed as a prediction.

Assumptions & free parameters 4 free parameters · 6 assumptions · 2 invented entities

The method depends on imported, unproven criteria and hand-set hyperparameters. The causal interpretation of the DII weights comes from the authors' own PNAS 2024 paper, not from a derivation in this manuscript. Several optimization hyperparameters (lambda, epoch count, mini-batch size, learning rate schedule) are fixed empirically without sensitivity analysis. The community extraction assumes an acyclic, unidirectional hierarchy. The only independent support is the synthetic validation, which uses known ground truth and intentionally added noise to satisfy faithfulness.

free parameters (4)
  • threshold epsilon = average of all weights in G; in Fig. 2 black star
    Main hyperparameter; links are kept when G_alpha_beta > epsilon. The reported graphs use the data-dependent average, and the paper shows a plateau but provides no principled rule for real data.
  • tau_max (maximum lag) = 1, 60, 30 for the three test systems
    Lagged distances are tested for tau=1..tau_max; chosen per system from autocorrelation heuristics; SM Fig. S2 shows tau=1 alone degrades detection.
  • lambda scale in DII = 0.1 * dA_i,j(k), k=5% of N
    Point-adaptive neighborhood scale, empirical prefactor 0.1; affects gradient behavior.
  • optimization schedule = 500 epochs, 20 mini-batches of 100, Adam lr=5e-3 cosine
    Hand-set training hyperparameters; no sensitivity analysis.
assumptions (6)
  • domain assumption Causal sufficiency: no unobserved common drivers of observed variables.
    Stated in Introduction and discussed in SM Section V; if violated, communities may be merged or spurious links appear.
  • domain assumption Faithfulness: every conditional independence corresponds to d-separation; noise is added to enforce it.
    Invoked in Results when adding noise, and SM Section I; without it, DII weights may not reflect causal structure.
  • ad hoc to paper Causal interpretation of DII weights: X causes Y iff optimal weight component w_X in Eq. (2) is nonzero.
    Imported from same-group ref. 19, not proven here; load-bearing for the entire pipeline (SM Eq. S5).
  • domain assumption Community graph is acyclic with unidirectional inter-community couplings.
    The algorithm extracts levels from minimal autonomous sets assuming a hierarchical acyclic structure; Discussion acknowledges bidirectional couplings as future work.
  • ad hoc to paper The DII optimization reaches the global minimum.
    No guarantees; relies on Adam, mini-batching, and point-adaptive lambda; local minima could create spurious weights.
  • domain assumption Samples are independent initial conditions drawn from a stationary trajectory.
    Uses N=2000 evenly spaced frames from one trajectory; stationarity is assumed.
invented entities (2)
  • Dynamical communities
    purpose: Group variables that mutually influence each other into a single node of a community causal graph
    Operational construct introduced by the paper; it is an output of the algorithm, not an entity with independent physical evidence.
  • Community causal graph
    purpose: Directed acyclic graph of communities ordered by autonomy levels
    New representation; its correctness is checked only against synthetic ground truths.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Linear scaling causal discovery from high-dimensional time series by dynamical community detection." pith.science (2026). https://pith.science/paper/36D5XPAX

@misc{pith2026250110886,
  author       = {Pith},
  title        = {Pith review of: Linear scaling causal discovery from high-dimensional time series by dynamical community detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/36D5XPAX}},
  note         = {Machine review of arXiv:2501.10886}
}
read the original abstract

Understanding which parts of a dynamical system cause each other is extremely relevant in fundamental and applied sciences. However, inferring causal links from observational data, namely without direct manipulations of the system, is still computationally challenging, especially if the data are high-dimensional. In this study we introduce a framework for constructing causal graphs from high-dimensional time series, whose computational cost scales linearly with the number of variables. The approach is based on the automatic identification of dynamical communities, groups of variables which mutually influence each other and can therefore be described as a single node in a causal graph. These communities are efficiently identified by optimizing the Information Imbalance, a statistical quantity that assigns a weight to each putative causal variable based on its information content relative to a target variable. The communities are then ordered starting from the fully autonomous ones, whose evolution is independent from all the others, to those that are progressively dependent on other communities, building in this manner a community causal graph. We demonstrate the computational efficiency and the accuracy of our approach on time-discrete and time-continuous dynamical systems including up to 80 variables.

Figures

Figures reproduced from arXiv: 2501.10886 by the authors.

Figure 1
Figure 1. FIG. 1. Illustration of the algorithm, using a 15-dimensional [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Outcome and performance of the algorithm in three different t [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

66 extracted references · 61 canonical work pages

  1. [19]

    A. B. Barrett, L. Barnett, and A. K. Seth, Multivariate granger causality and generalized variance, Phys. Rev. E 81, 041907 (2010)

    work page 2010

  2. [1]

    Runge, A

    J. Runge, A. Gerhardus, G. Varando, V. Eyring, and G. Camps-Valls, Causal inference for time series, Nature Reviews Earth & Environment 4 (2023)

    work page 2023

  3. [2]

    To level out the fluctuation ranges of different variables, we first scale each variable by its standard de- viation over the entire trajectory

    between a distance measure built with all dynamic variables at time t = 0 and a distance built with a single variable X β at time t = τ . To level out the fluctuation ranges of different variables, we first scale each variable by its standard de- viation over the entire trajectory. Thus, the weights ob- tained by this optimization are ˆwβ = argmin w DII ( dw...

    work page

  4. [3]

    Following the intuition that different couplings might manifest at different time scales [ 19–21] we repeat the optimization in Eq

    by stating that X α is a direct or indirect cause of X β when the α-component of ˆwβ, denoted by ˆwα β , is nonzero. Following the intuition that different couplings might manifest at different time scales [ 19–21] we repeat the optimization in Eq. ( 5) for several values of τ between 1 and τmax, where τmax is a hyper parameter which, in applications, we ta...

    work page

  5. [4]

    microscopic

    tends to Eq. ( 1) in the limit λ → 0. If distance dA depends on a set of parameters w, this for- mulation allows optimizing such parameters by gradient descent. Our algorithm, illustrated in Fig. 1, can be conceptu- ally divided into three parts: in part i) the DII is mini- mized to infer each variable’s autonomous set, namely the set of variables that di...

    work page

  6. [5]

    Spirtes, C

    P. Spirtes, C. Glymour, and R. Scheines, Causation, Pre- diction, and Search (The MIT Press, 2001)

    work page 2001

  7. [6]

    In the SM (Fig

    as a valuable feature of our approach. In the SM (Fig. S2) we show that using only τ = 1 would significantly decrease the detection power of the algorithm. The codes implementing our algorithm are available in the Python library DADApy [ 31]. Acknowledgments — This work was partially funded by NextGenerationEU through the Italian National Cen- tre for HPC ...

    work page

  8. [7]

    J. Runge, Causal network reconstruction from time se- ries: From theoretical assumptions to practical estima- tion, Chaos: An Interdisciplinary Journal of Nonlinear Science 28, 075310 (2018)

    work page 2018

Show all 66 references
  1. [10]

    Spirtes and C

    P. Spirtes and C. Glymour, An algorithm for fast recovery of sparse causal graphs, Social Science Computer Review 9, 62 (1991)

    1991

  2. [11]

    Geiger, T

    D. Geiger, T. Verma, and J. Pearl, Identifying indepen- dence in bayesian networks, Networks 20, 507 (1990)

    1990

  3. [12]

    Verma and J

    T. Verma and J. Pearl, Causal networks: Semantics and expressiveness, in Machine Intelligence and Pattern Recognition, Machine Intelligence and Pattern Recogni- tion No. C (1990) pp. 69–76, c ed

    1990

  4. [14]

    Verma and J

    T. Verma and J. Pearl, Equivalence and synthesis of causal models, in Proceedings of the Sixth Annual Con- ference on Uncertainty in Artificial Intelligence , UAI ’90 (Elsevier Science Inc., USA, 1990) p. 255–270

    1990

  5. [15]

    K. J. Blinowska, R. Ku´ s, and M. Kami´ nski, Granger causality and information flow in multivariate processes, Phys. Rev. E 70, 050902 (2004) . 6

    2004

  6. [16]

    Runge, P

    J. Runge, P. Nowack, M. Kretschmer, S. Flaxman, and D. Sejdinovic, Detecting and quantifying causal associa- tions in large nonlinear time series datasets, Science Ad- vances 5, eaau4996 (2019)

    2019

  7. [17]

    J. Runge, Discovering contemporaneous and lagged causal relations in autocorrelated nonlinear time series datasets, in Proceedings of the 36th Conference on Un- certainty in Artificial Intelligence (UAI) , Proceedings of Machine Learning Research, Vol. 124, edited by J. Peters ...

    2020

  8. [18]

    C. W. J. Granger, Investigating causal relations by econo- metric models and cross-spectral methods, Econometrica 37, 424–438 (1969)

    1969

  9. [20]

    Tirabassi, R

    G. Tirabassi, R. Sevilla-Escoboza, J. M. Buld´ u, and C. Masoller, Inferring the connectivity of coupled oscil- lators from time-series statistical similarity analysis, Sci. Rep. 5, 10829 (2015)

    2015

  10. [22]

    multi- body

    (Figs. 2d-f), and two 40-dimensional Lorenz 96 sys- tems [ 23] unidirectionally coupled (Figs. 2g-i). Since de- terministic relationships are known to violate a condition known as faithfulness [ 2, 8] we added a small white noise to each variable while integrating the dynamic ...

    2000

  11. [23]

    Glielmo, C

    A. Glielmo, C. Zeni, B. Cheng, G. Cs´ anyi, and A. Laio, Ranking the information content of distance measures, PNAS Nexus 1, pgac039 (2022), https://academic.oup.com/pnasnexus/article- pdf/1/2/pgac039/53406126/pgac039.pdf

    2022

  12. [24]

    R. Wild, F. Wodaczek, V. Del Tatto, B. Cheng, and A. Laio, Automatic feature selection and weighting in molecular systems using differentiable information imbal- ance, Nature Communications 16, 270 (2025)

    2025

  13. [25]

    Del Tatto, G

    V. Del Tatto, G. Fortunato, D. Bueti, and A. Laio, Ro- bust inference of causality in high-dimensional dynami- cal processes from the information imbalance of distance ranks, Proceedings of the National Academy of Sciences 121, e2317256121 (2024)

    2024

  14. [26]

    D. A. Smirnov, Quantification of causal couplings via dy- namical effects: A unifying perspective, Phys. Rev. E 90, 062921 (2014)

    2014

  15. [27]

    E. N. Lorenz, Deterministic nonperiodic flow, Journal of Atmospheric Sciences 20, 130 (1963)

    1963

  16. [28]

    E. N. Lorenz, Predictability – a problem partly solved, in Predictability of Weather and Climate , edited by T. Palmer and R. Hagedorn (Cambridge University Press, 2006) p. 40–58

    2006

  17. [29]

    N. X. Vinh, J. Epps, and J. Bailey, Information theoretic measures for clusterings comparison: is a correction for chance necessary?, in Proceedings of the 26th Annual In- ternational Conference on Machine Learning , ICML ’09 (Association for Computing Machinery, New York, NY,...

    2009

  18. [30]

    J. Wahl, U. Ninad, and J. Runge, Foundations of causal discovery on groups of variables, Journal of Causal Infer- ence 12, 20230041 (2024)

    2024

  19. [31]

    S. E. Schaeffer, Graph clustering, Computer Science Re- view 1, 27 (2007)

    2007

  20. [32]

    Tononi, A

    G. Tononi, A. R. McIntosh, D. Russell, and G. M. Edel- man, Functional clustering: Identifying strongly interac- tive brain regions in neuroimaging data, NeuroImage 7, 133 (1998)

    1998

  21. [33]

    Villani, A

    M. Villani, A. Roli, A. Filisetti, M. Fiorucci, I. Poli, and R. Serra, The search for candidate relevant subsets of variables in complex systems, Artificial Life 21, 412 (2015)

    2015

  22. [34]

    D’Addese, L

    G. D’Addese, L. Sani, L. La Rocca, R. Serra, and M. Vil- lani, Asymptotic information-theoretic detection of dy- namical organization in complex systems, Entropy 23, 10.3390/e23040398 (2021)

    2021 doi

  23. [35]

    D’Addese, M

    G. D’Addese, M. Casari, R. Serra, and M. Vil- lani, A fast and effective method to identify relevant sets of variables in complex systems, Mathematics 9, 10.3390/math9091022 (2021)

    2021 doi

  24. [36]

    Glielmo, I

    A. Glielmo, I. Macocco, D. Doimo, M. Carli, C. Zeni, R. Wild, M. d’Errico, A. Rodriguez, and A. Laio, Dadapy: Distance-based analysis of data-manifolds in python, Patterns 3, 100589 (2022)

    2022

  25. [37]

    direct” if there is at least a pair of variables, X α ∈ A and X β ∈ B , such that X α is a direct cause of X β. In contrast, we call such a link “indirect

    P. Virtanen, R. Gommers, T. E. Oliphant, M. Haber- land, T. Reddy, D. Cournapeau, E. Burovski, P. Pe- terson, W. Weckesser, J. Bright, S. J. van der Walt, M. Brett, J. Wilson, K. J. Millman, N. Mayorov, A. R. J. Nelson, E. Jones, R. Kern, E. Larson, C. J. Carey, ˙I. Po- lat, Y...

    2020

  26. [38]

    can be skipped when- ever the all-variable adjacency matrix constructed from matrix Gαβ (step i) of the algorithm) does not display any link between variables in A and variables in B, which is a necessary condition to have a direct link from A to B. In the SM we validate this ...

  27. [39]

    Logistic maps The equations of the noisy coupled logistic maps shown in Figs. 1 and 2a of the main text, structured in three communities with five variables each ( µ, ν ∈ { 0, 1, 2}; α, β ∈ {0, 1, 2, 3, 4}), read: X α µ (t + 1) = X α µ (t) · ( rα µ − rα µ · X α µ (t) − ∑ 4 β=0 ...

  28. [40]

    (10) The coupling strength was fixed to c = 0 .3 and dνµ, which defines the interaction topology among communi- ties (Fig

    Lorenz systems The system of 5 coupled Lorenz systems ( µ, ν ∈ {0, 1, 2, 3, 4}) is described by the following Itˆ o stochastic differential equations, reported here for a single system (or community) µ:          dX 0 µ = 10 (X 1 µ − X 0 µ)dt + dW 0 µ dX 1 µ = (X 0 µ(28...

  29. [41]

    The trajectory was calculated with a sampling time of δt = 0.003

    were integrated using the Euler-Maruyama al- gorithm for Itˆ o equations. The trajectory was calculated with a sampling time of δt = 0.003. The noise was taken independent for each variable at each time step (autocor- relation A(τ ) = δ(τ ))

  30. [42]

    Lorenz 96 systems The two unidirectionally coupled Lorenz 96 systems of 40 variables each ( µ, ν ∈ { 0, 1}, α, β ∈ { 0, 1, ..., 39}) are defined by the following ordinary differential equations: dX α µ = ( (X α+1 µ − X α−2 µ )X α−1 µ − X α µ + Fµ + +c δµ,1 X α 0 ) dt + dW α µ , ...

  31. [43]

    struct ural causal model

    were integraterd with time step dt = 0.03 using uncorrelated noise and Euler-Maruyama algorithm as above. In the Lorenz and Lorenz 96 systems, the coupling terms among communities have the same functional form employed in ref. [ 19]. Appendix C: Validation measures — The adjus...

    2000

  32. [44]

    J. Runge, Causal network reconstruction from time series: From t heoretical assumptions to practical estimation, Chaos: An Interdisciplinary Journal of Nonlinear Science 28, 075310 (2018)

    2018

  33. [45]

    Runge, A

    J. Runge, A. Gerhardus, G. Varando, V. Eyring, and G. Camps-Va lls, Causal inference for time series, Nature Reviews Earth & Environment 4 (2023)

    2023

  34. [46]

    Spirtes, C

    P. Spirtes, C. Glymour, and R. Scheines, Causation, Prediction, and Search (The MIT Press, 2001)

    2001

  35. [47]

    C. K. Assaad, E. Devijver, and E. Gaussier, Survey and evaluati on of causal discovery methods for time series, J. Artif. Int. Res. 73, 10.1613/jair.1.13428 (2022)

    2022 doi

  36. [48]

    J. Runge, Discovering contemporaneous and lagged causal rel ations in autocorrelated nonlinear time series datasets, in Proceedings of the 36th Conference on Uncertainty in Artific ial Intelligence (UAI) , Proceedings of Machine Learning Research, Vol. 124, edited by J. Peters ...

    2020

  37. [49]

    D. M. Hausman and J. Woodward, Independence, invariance and the causal markov condition, British Journal for the Philosophy of Science 50, 521–583 (1999)

    1999

  38. [50]

    Pearl, Causality: Models, Reasoning and Inference , 2nd ed

    J. Pearl, Causality: Models, Reasoning and Inference , 2nd ed. (Cambridge University Press, USA, 2009)

    2009

  39. [51]

    Chicharro and S

    D. Chicharro and S. Panzeri, Algorithms of causal inference for th e analysis of effective connectivity among brain regions, Frontiers in Neuroinformatics 8, 10.3389/fninf.2014.00064 (2014)

    2014

  40. [52]

    Geiger, T

    D. Geiger, T. Verma, and J. Pearl, Identifying independence in bayesian networks, Networks 20, 507 (1990)

    1990

  41. [53]

    Verma and J

    T. Verma and J. Pearl, Causal networks: Semantics and expres siveness, in Machine Intelligence and Pattern Recognition , Machine Intelligence and Pattern Recognition No. C (1990) pp . 69–76, c ed

    1990

  42. [54]

    Runge, P

    J. Runge, P. Nowack, M. Kretschmer, S. Flaxman, and D. Sejdin ovic, Detecting and quantifying causal associations in large nonlinear time series datasets, Science Advances 5, eaau4996 (2019) . 11

    2019

  43. [55]

    Spirtes and C

    P. Spirtes and C. Glymour, An algorithm for fast recovery of spars e causal graphs, Social Science Computer Review 9, 62 (1991)

    1991

  44. [56]

    Verma and J

    T. Verma and J. Pearl, Equivalence and synthesis of causal models, in Proceedings of the Sixth Annual Conference on Uncertainty in Artificial Intelligence , UAI ’90 (Elsevier Science Inc., USA, 1990) p. 255–270

    1990

  45. [57]

    C. W. J. Granger, Investigating causal relations by econom etric models and cross-spectral methods, Econometrica 37, 424–438 (1969)

    1969

  46. [58]

    A. B. Barrett, L. Barnett, and A. K. Seth, Multivariate grange r causality and generalized variance, Phys. Rev. E 81, 041907 (2010)

    2010

  47. [59]

    K. J. Blinowska, R. Ku´ s, and M. Kami´ nski, Granger causa lity and information flow in multivariate processes, Phys. Rev. E 70, 050902 (2004)

    2004

  48. [60]

    Schreiber, Measuring information transfer, Phys

    T. Schreiber, Measuring information transfer, Phys. Rev. Lett. 85, 461 (2000)

    2000

  49. [61]

    Del Tatto, G

    V. Del Tatto, G. Fortunato, D. Bueti, and A. Laio, Robust infe rence of causality in high-dimensional dynamical processes from the information imbalance of distance ranks, Proceedings of the National Academy of Sciences 121, e2317256121 (2024)

    2024

  50. [62]

    Paluˇ s, V

    M. Paluˇ s, V. Kom´ arek, Z. c. Hrn ˇ c ´ ıˇ r, and K.ˇStˇ erbov´ a, Synchronization as adjustment of information rates: Detection from bivariate time series, Phys. Rev. E 63, 046211 (2001)

    2001

  51. [63]

    Paluˇ s and M

    M. Paluˇ s and M. Vejmelka, Directionality of coupling from bivariate time series: How to avoid false causalities and misse d connections, Phys. Rev. E 75, 056211 (2007)

    2007

  52. [64]

    Runge, tigramite (2017)

    J. Runge, tigramite (2017)

    2017

  53. [65]

    J. Runge, Conditional independence testing based on a n earest-neighbor estimator of conditional mutual information , in Proceedings of the Twenty-First International Conference on Artificial Intelligence and Statistics , Proceedings of Machine Learning Research, Vol. 84, edite...

    2018

  54. [66]

    Reiter, A

    N.-D. Reiter, A. Gerhardus, J. Wahl, and J. Runge, Causal infe rence on process graphs, part i: The structural equation process representation (2023)

    2023

  55. [67]

    Glielmo, I

    A. Glielmo, I. Macocco, D. Doimo, M. Carli, C. Zeni, R. Wild, M. d’Errico, A. Rodriguez, and A. Laio, Dadapy: Distance- based analysis of data-manifolds in python, Patterns 3, 100589 (2022)

    2022

  56. [68]

    Bradbury, R

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

    2018

  57. [69]

    R. Wild, F. Wodaczek, V. Del Tatto, B. Cheng, and A. Laio, A utomatic feature selection and weighting in molecular systems using differentiable information imbalance, Nature Communications 16, 270 (2025)

    2025

  58. [70]

    D. P. Kingma and J. Ba, Adam: A method for stochastic optimi zation (2017), arXiv:1412.6980 [cs.LG]

    2017 arXiv

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.