Inferring causal networks of dynamical systems through transient dynamics and perturbation.

  title={Inferring causal networks of dynamical systems through transient dynamics and perturbation.},
  author={George Stepaniants and Bingni W. Brunton and J. Nathan Kutz},
  journal={Physical review. E},
  volume={102 4-1},
Inferring causal relations from time series measurements is an ill-posed mathematical problem, where typically an infinite number of potential solutions can reproduce the given data. We explore in depth a strategy to disambiguate between possible underlying causal networks by perturbing the network, where the forcings are either targeted or applied at random. The resulting transient dynamics provide the critical information necessary to infer causality. Two methods are shown to provide accurate… 

Figures from this paper

Data-driven causal analysis of observational time series in ecology

This work provides an accessible and critical review of three statistical causal inference approaches popular in ecological time series analysis: pairwise correlation, Granger causality, and state space reconstruction.

Network inference via process motifs for lagged correlation in linear stochastic processes

This work formulate two PEMs that correct for confounding factors and for reverse causation, and provide promising alternatives to current paradigms for the inference of linear models from time-series data, including Granger causality, vector-autoregression, and sparse inverse covariance estimation.

Backpropagation on Dynamical Networks

A network inference method based on the backpropagation through time (BPTT) algorithm commonly used to train recurrent neural networks that was found to be succesful in identifying the connectivity structure for Lorenz, Chua and FitzHugh-Nagumo oscillators.

Swarm Modeling With Dynamic Mode Decomposition

This work applies a purely data-driven method to learn local interactions of homogeneous swarms through observation data and to generate similar swarming behaviour using the learned model, called swarmDMD, which can faithfully reconstruct the swarm dynamics.

Swarm Modelling with Dynamic Mode Decomposition

The proposed swarmDMD approach will be useful for studying multi-agent systems found in biology, physics, and engineering and provide a comprehensive analysis on the efficacy of different observation data types on the modelling, where it is shown that inter-agent distance yields the most accurate models.

Embedding entropy: a nonlinear measure of dynamical causality

A unified mathematical framework for the so-called dynamical causality (DC), which can detect causal interactions over time is presented; notably, this framework covers Granger causality, transfer entropy, embedding causality and their conditional versions.

Identification of Couplings in Adaptive Dynamical Networks of Time-Delayed Feedback Oscillators

An approach to solve the inverse problem of the reconstruction of the network of time-delay oscillators from their time series is proposed and studied in the case of the nonstationary connectivity



Inferring connectivity in networked dynamical systems: Challenges using Granger causality.

The results show a significant systematic disparity between the original and inferred network, unless the true structure is extremely sparse or dense, and advocate the need to perform comparisons with any network inference method since the inferred connectivity results appear to have very little to do with the ground truth system.

Network inference using directed information: The deterministic limit

This paper proposes a remedy called restricted directed information (RDI), and shows that, RDI recovers the graph correctly in all instances, deterministic or stochastic, where there is enough information to recover the graph uniquely.

Advances to Bayesian network inference for generating causal networks from observational biological data

A novel influence score for DBNs is developed that attempts to estimate both the sign (activation or repression) and relative magnitude of interactions among variables and reduces a significant portion of false positive interactions in the recovered networks.

Detecting Causality in Complex Ecosystems

A new method, based on nonlinear state space reconstruction, that can distinguish causality from correlation is introduced, and extends to nonseparable weakly connected dynamic systems (cases not covered by the current Granger causality paradigm).

Systematic errors in connectivity inferred from activity in strongly recurrent networks

It is demonstrated that strongly recurrent circuits inferred from neural activity, even with unlimited data from every neuron, are biased and Synapses are inferred between unconnected but correlated neurons.

Revealing network connectivity from response dynamics.

  • M. Timme
  • Mathematics
    Physical review letters
  • 2007
This work considers networks of coupled phase oscillators and explicitly study their long-term stationary response to temporally constant driving, finding good predictions of the actual connectivity even for formally underdetermined problems.

Methods for causal inference from gene perturbation experiments and validation

The results suggest that prediction and prioritization of future experimental interventions, such as gene deletions, can be improved by using the statistical inference techniques used in this work.