Selected publications

LiNGAM: Linear Non-Gaussian Acyclic Models

Standard LiNGAM

S. Shimizu, P. O. Hoyer, A. Hyvärinen, and A. Kerminen. A linear non-gaussian acyclic model for causal discovery. Journal of Machine Learning Research, 7: 2003--2030, 2006.
pdf erratum
LiNGAM homepage
Matlab/Octave code
A short introduction
[Proposes a new statistical method to discover linear non-gaussian acyclic causal models of continuous variables. This is based on UAI2005 paper.]

New S. Shimizu, A. Hyvärinen, Y. Kawahara and T. Washio. A direct method for estimating a causal ordering in a linear non-Gaussian acyclic model. In Proc. 25th Conf. on Uncertainty in Artificial Intelligence (UAI2009), Montreal, Canada, 2009.
pdf notes
A short introduction erratum
(A Matlab implementation of this algorithm will be made available probably soon.)
[Proposes a new estimation algorithm for LiNGAM. The new method requires no algorithmic parameters and is guaranteed to converge to the right solution within a small ﬁxed number of steps if the data strictly follows the model.]

P. O. Hoyer, A. Hyvärinen, R. Scheines, P. Spirtes, J. Ramsey, G. Lacerda and S. Shimizu. Causal discovery of linear acyclic models with arbitrary distributions. In Proc. 24th Conf. on Uncertainty in Artificial Intelligence (UAI2008), pp. 282--289, Helsinki, Finland, 2008.
pdf
[Shows how to combine the conditional-independence approach and LiNGAM so that it works for both gaussian and non-gaussian variables.]

Time series

A. Hyvärinen, S. Shimizu, and P. O. Hoyer. Causal modelling combining instantaneous and lagged effects: an identifiable model based on non-Gaussianity. In Proc. Int. Conf. on Machine Learning (ICML2008), pp. 424--431, Helsinki, Finland, 2008.
pdf
Videolecture
[Shows how to combine LiNGAM and autoregressive models.]

Latent variables

S. Shimizu, P. O. Hoyer and A. Hyvärinen. Estimation of linear non-Gaussian acyclic models for latent factors. Neurocomputing, 72: 2024-2027, 2009.
pdf
[Proposes that a non-gaussian assumption allows linear acyclic models for latent factors uniquely identified under standard assumptions.]

S. Shimizu and A. Hyvärinen. Discovery of linear non-gaussian acyclic models in the presence of latent classes. In Proc. 14th Int. Conf. on Neural Information Processing (ICONIP2007), pp. 752--761 Kitakyushu, Japan, 2008.
pdf
[Extends the preceding method above to cases where latent classes are present. The new method finds hidden groups of samples that have similiar DAG structures.]

P. O. Hoyer, S. Shimizu, A. Kerminen, and M. Palviainen. Estimation of causal effects using linear non-gaussian causal models with hidden variables. International Journal of Approximate Reasoning, 49(2): 362-378, 2008.
pdf (7.0MB)
[Discusses an extension of the preceding method above in the presence of unobserved confounding (continuous) variables.]

Reliability analysis

New Y. Komatsu, S. Shimizu and H. Shimodaira. Computing p-values of LiNGAM outputs via Multiscale Bootstrap. arXiv:0909.2904v1, Sep, 2009.
pdf
[Proposes a method to evaluate reliability of variable orderings estimated by LiNGAM. ]

Structural equation modeling

S. Shimizu and Y. Kano. Use of non-normality in structural equation modeling: Application to direction of causation. Journal of Statistical Planning and Inference, 138: 3483--3491, 2008.
pdf
[Proposes a non-Gaussian method to obtain an empirical evidence of the direction of a path between two variables.]

Independent component analysis

A. Hyvärinen and S. Shimizu. A quasi-stochastic gradient algorithm for variance-dependent component analysis. In Proc. International Conference on Artificial Neural Networks (ICANN2006), pp.211--220, Athens, Greece, 2006.
pdf
[Proposes a new algorithm for a generalized least squares approach using higher-order moments for estimating variance dependent component analysis model, a variant of ICA. The method proposed here does not require time structured signals nor parametric models.]

S. Shimizu, A. Hyvärinen, Y. Kano, P. O. Hoyer, and A. Kerminen. Testing significance of mixing and demixing coefficients in ICA. In Proc. International Conference on Independent Component Analysis and Blind Signal Separation (ICA2006), Charleston, SC, USA, pp.901--908, 2006.
pdf
Matlab code
[Provides test statistics to examine significance of mixing and demixing coefficients estimated by FastICA.]

Y. Kano, Y. Miyamoto, and S. Shimizu. Factor rotation and ICA. In Proc. International Conference on Independent Component Analysis and Blind Signal Separation (ICA2003), Nara, Japan, pp.101--105, 2003.
pdf
[Points out a connection between independent component analysis and traditional factor analysis. The key of the connection is "factor rotation".]

Data mining: misc.

S. Shimizu, T. Washio, A. Hyvärinen, and S. Imoto. Finding exogenous variables in data with many more variables than observations. arXiv:0904.0838v1, April, 2009.
pdf
[Proposes a method to find exogenous variables in a linear non-Gaussian acyclic model with some restriction on error terms. The restriction might be rather strong, but probably not very much in some applications. ]

Shohei Shimizu