Causal deconvolution by algorithmic generative models


Complex behaviour emerges from interactions between objects produced by different generating mechanisms. Yet to decode their causal origin(s) from observations remains one of the most fundamental challenges in science. Here we introduce a universal, unsupervised and parameter-free model-oriented approach, based on the seminal concept and the first principles of algorithmic probability, to decompose an observation into its most likely algorithmic generative models. Our approach uses a perturbation-based causal calculus to infer model representations. We demonstrate its ability to deconvolve interacting mechanisms regardless of whether the resultant objects are bit strings, space–time evolution diagrams, images or networks. Although this is mostly a conceptual contribution and an algorithmic framework, we also provide numerical evidence evaluating the ability of our methods to extract models from data produced by discrete dynamical systems such as cellular automata and complex networks. We think that these separating techniques can contribute to tackling the challenge of causation, thus complementing statistically oriented approaches.

A preprint version of the article is available at ArXiv.

Access options

Rent or Buy article

Get time limited or full article access on ReadCube.


All prices are NET prices.

Fig. 1: Proof of concept applied to a binary string composed of two segments with different underlying generating mechanisms (computer programs).
Fig. 2: Training-free separation of intertwined programs despite their statistical similarity from an observer’s perspective.
Fig. 3: Algorithmic similarity and graph hierarchical decomposition leading to causal clustering.
Fig. 4: Unsupervised graph deconvolution identifies each different topological generating mechanism.

Data availability

The data that support the plots within this paper are available from the corresponding author upon request.


  1. 1.

    Zenil, H. et al. An algorithmic information calculus for causal discovery and reprogramming systems. Preprint at (2018).

  2. 2.

    Zenil, H., Kiani, N. A., Zea, A. A., Rueda-Toicen, A. & Tegnér, J. Data dimension reduction and network sparsification based on minimal algorithmic information loss. Preprint at (2018).

  3. 3.

    Lloyd, S. P. Least squares quantization in PCM. IEEE Trans. Inform. Theory 28, 129–137 (1982).

    MathSciNet  Article  Google Scholar 

  4. 4.

    Kaufman, L. & Rousseeuw, P. J. in Statistical Data Analysis Based on the L1-Norm and Related Methods (North-Holland, Amsterdam, 1987).

    Google Scholar 

  5. 5.

    Ben-Hur, A., Horn, D., Siegelmann, H. & Vapnik, V. N. Support vector clustering. J. Mach. Learn. Res. 2, 125–137 (2001).

    MATH  Google Scholar 

  6. 6.

    Milo, R. et al. Network motifs: simple building blocks of complex networks. Science 298, 824–827 (2002).

    Article  Google Scholar 

  7. 7.

    Newman, M. E. J. Finding community structure in networks using the eigenvectors of matrices. Phys. Rev. E 74, 036104 (2006).

    MathSciNet  Article  Google Scholar 

  8. 8.

    Benczur, A. & Karger, D. R. Approximating s-t minimum cuts in O(n 2)-time. In Proc. Twenty-Eighth Annual ACM Symposium on the Theory of Computing 47–55 (ACM, 1996).

  9. 9.

    Spielman, D. A. & Srivastava, N. Graph sparsification by effective resistances. In Proc. Fortieth Annual ACM Symposium on Theory of Computing 563–568 (ACM, 2008).

  10. 10.

    Spielman, D. A. & Teng, S.-H. Spectral sparsification of graphs. SIAM J. Comput. 40, 981–1025 (2011).

    MathSciNet  Article  Google Scholar 

  11. 11.

    Liu, M., Liu, B. & Wei, F. Graphs determined by their (signless) Laplacian spectra. Electron. J. Linear Algebra 22, 112–124 (2011).

    MathSciNet  MATH  Google Scholar 

  12. 12.

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

    Article  Google Scholar 

  13. 13.

    Schreiber, T. Measuring information transfer. Phys. Rev. Lett. 85, 461–464 (2000).

    Article  Google Scholar 

  14. 14.

    Pearl, J. Causality: Models, Reasoning and Inference (Cambridge University Press, Cambridge, 2000).

    Google Scholar 

  15. 15.

    Solomonoff, R. J. A formal theory of inductive inference: parts 1 and 2. Inform. Control 7, 1–22–224–254 (1964).

    MathSciNet  MATH  Google Scholar 

  16. 16.

    Watanabe, S. in Frontiers of Pattern Recognition (ed. Watanabe, S.) 561–568 (Academic Press, New York, 1972).

  17. 17.

    Williams, P. L. & Beer, R. D. Nonnegative decomposition of multivariate information. Preprint at (2010).

  18. 18.

    Lizier, J. T., Bertschinger, N., Jost, J. & Wibral, M. Information decomposition of target effects from multi-source interactions: perspectives on previous, current and future work. Entropy 20, 307 (2018).

    Article  Google Scholar 

  19. 19.

    Li, M. & Vitányi, P. M. B. An Introduction to Kolmogorov Complexity and Its Applications 3rd edn (Springer, New York, 2009).

  20. 20.

    Li, M., Chen, X., Li, X., Ma, B. & Vitányi, P. M. B. The similarity metric. IEEE Trans. Inf. Theory 50, 3250–3264 (2004).

    MathSciNet  Article  Google Scholar 

  21. 21.

    Bennett, C. H., Gács, P., Li, M., Vitányi, P. M. B. & Zurek, W. H. Information distance. IEEE Trans. Inf. Theory 44, 1407–1423 (1998).

    MathSciNet  Article  Google Scholar 

  22. 22.

    Cilibrasi, R. & Vitanyi, P. M. B. Clustering by compression. IEEE Trans. Inf. Theory 51, 1523–1545 (2005).

    MathSciNet  Article  Google Scholar 

  23. 23.

    Shannon, C. E. A mathematical theory of communication. Bell Syst. Tech. J. 27, 379–423 (1948).

    MathSciNet  Article  Google Scholar 

  24. 24.

    Ince, R. A. A. Measuring multivariate redundant information with pointwise common change in surprisal. Entropy 19, 318 (2017).

    Article  Google Scholar 

  25. 25.

    Strelioff, C. C. & Crutchfield, J. P. Bayesian structural inference for hidden processes. Phys. Rev. E 89, 042119 (2014).

    Article  Google Scholar 

  26. 26.

    Shalizi, C. R. & Crutchfield, J. P. Computational mechanics: pattern and prediction, structure and simplicity. J. Stat. Phys. 104, 819–881 (2001).

    MathSciNet  Article  Google Scholar 

  27. 27.

    Delahaye, J.-P. & Zenil, H. Numerical evaluation of the complexity of short strings: a glance into the innermost structure of algorithmic randomness. Appl. Math. Comput. 219, 63–77 (2012).

    MATH  Google Scholar 

  28. 28.

    Soler-Toscano, F., Zenil, H., Delahaye, J.-P. & Gauvrit, N. Calculating Kolmogorov complexity from the frequency output distributions of small Turing machines. PLoS ONE 9, e96223 (2014).

    Article  Google Scholar 

  29. 29.

    Hutter, M. Universal Artificial Intelligence (EATCS Series, Springer, Berlin, 2005).

  30. 30.

    Gauvrit, N., Zenil, H. & Tegnér, J. in Representation and Reality: Humans, Animals and Machines (eds Dodig-Crnkovic, G. & Giovagnoli, R.) 117–139 (Springer, Berlin,Berlin, 2017).

  31. 31.

    Rissanen, J. Modeling by shortest data description. Automatica 14, 465–658 (1978).

    Article  Google Scholar 

  32. 32.

    Levin, L. A. Universal search problems. Probl. Inform. Transm. 9, 265–266 (1973).

    Google Scholar 

  33. 33.

    Schmidhuber, J. The speed prior: a new simplicity measure yielding, near-optimal computable predictions. In Proc. 15th annual conference on Computational Learning Theory (COLT 2002) (eds Kivinen, J. & Sloan, R. H.) 216–228 (Springer, Sydney, 2002).

  34. 34.

    Daley, R. P. Minimal-program complexity of pseudo-recursive and pseudo-random sequences. Math. Syst. Theory 9, 83–94 (1975).

    MathSciNet  Article  Google Scholar 

  35. 35.

    Zenil, H., Badillo, L., Hernández-Orozco, S. & Hernández-Quiroz, F. Coding-theorem like behaviour and emergence of the universal distribution from resource-bounded algorithmic probability. Int. J. Parallel Emergent Distrib. Syst. (2018).

  36. 36.

    Hernández-Orallo, J. Computational measures of information gain and reinforcement in inference processes. AI Commun. 13, 49–50 (2000).

    Google Scholar 

  37. 37.

    Hernández-Orallo, J. Universal and cognitive notions of part. In Proc. 4th Systems Science European Congress 711–722 (EC, 1999).

  38. 38.

    Solomonoff, R. J. The time scale of artificial intelligence: reflections on social effects. Human. Syst. Manag. 5, 149–153 (1985).

    Google Scholar 

  39. 39.

    Zenil, H. et al. A decomposition method for global evaluation of Shannon entropy and local estimations of algorithmic complexity. Entropy 20, 605 (2018).

    Article  Google Scholar 

  40. 40.

    Chaitin, G. J. On the length of programs for computing finite binary sequences. J. ACM 13, 547–569 (1966).

    MathSciNet  Article  Google Scholar 

  41. 41.

    Levin, L. A. Laws of information conservation (non-growth) and aspects of the foundation of probability theory. Probl. Inf. Transm. 10, 206–210 (1974).

    Google Scholar 

  42. 42.

    Zenil, H., Kiani, N. A. & Tegnér, J. Symmetry and correspondence of algorithmic complexity over geometric, spatial and topological representations. Entropy 20, 534 (2018).

    Article  Google Scholar 

  43. 43.

    Zenil, H., Soler-Toscano, F., Delahaye, J.-P. & Gauvrit, N. Two-dimensional Kolmogorov complexity and validation of the coding theorem method by compressibility. PeerJ Comput. Sci. 1, e23 (2013).

    Article  Google Scholar 

  44. 44.

    Riedel, J. & Zenil, H. Rule primality and compositional emergence of Turing-universality from elementary cellular automata. J. Cell. Autom. 13, 479–497 (2018).

    MathSciNet  Google Scholar 

  45. 45.

    Pearl, J. To build truly intelligent machines, teach them cause and effect. Quanta Magazine (15 May 2018).

  46. 46.

    Minsky, M. The limits of understanding. World Science Festival

Download references


H.Z. was supported by Swedish Research Council (Vetenskapsrådet) grant number 2015-05299. J.T. was supported by the King Abdullah University of Science and Technology.

Author information




H.Z., N.A.K. and J.T. conceived and designed the algorithms. H.Z. designed the experiments and carried out the calculations and numerical experiments. A.A.Z. and H.Z. conceived the online tool to illustrate the method applied to simple examples based on this paper. All authors contributed to the writing of the paper.

Corresponding authors

Correspondence to Hector Zenil or Narsis A. Kiani or Jesper Tegnér.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Supplementary Methods, Figures and References.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Zenil, H., Kiani, N.A., Zea, A.A. et al. Causal deconvolution by algorithmic generative models. Nat Mach Intell 1, 58–66 (2019).

Download citation

Further reading