Learning Latent Variable Gaussian Graphical Models

Gaussian graphical models (GGM) have been widely used in many high-dimensional applications ranging from biological and financial data to recommender systems. Sparsity in GGM plays a central role both statistically and computationally. Unfortunately, real-world data often does not fit well to sparse graphical models. In this paper, we focus on a family of latent variable Gaussian graphical models (LVGGM), where the model is conditionally sparse given latent variables, but marginally non-sparse. In LVGGM, the inverse covariance matrix has a low-rank plus sparse structure, and can be learned in a regularized maximum likelihood framework. We derive novel parameter estimation error bounds for LVGGM under mild conditions in the high-dimensional setting. These results complement the existing theory on the structural learning, and open up new possibilities of using LVGGM for statistical inference.

[1]  Neil D. Lawrence,et al.  Residual Component Analysis: Generalising PCA for more flexible inference in linear-Gaussian models , 2012, ICML 2012.

[2]  Pradeep Ravikumar,et al.  Dirty Statistical Models , 2013, NIPS.

[3]  Karim Lounici High-dimensional covariance matrix estimation with missing observations , 2012, 1201.2577.

[4]  Alfred O. Hero,et al.  Robust Shrinkage Estimation of High-Dimensional Covariance Matrices , 2010, IEEE Transactions on Signal Processing.

[5]  Jonathan E. Taylor,et al.  On model selection consistency of penalized M-estimators: a geometric theory , 2013, NIPS.

[6]  Martin J. Wainwright,et al.  A unified framework for high-dimensional analysis of $M$-estimators with decomposable regularizers , 2009, NIPS.

[7]  Pradeep Ravikumar,et al.  BIG & QUIC: Sparse Inverse Covariance Estimation for a Million Variables , 2013, NIPS.

[8]  Dmitry M. Malioutov,et al.  Walk-Sums and Belief Propagation in Gaussian Graphical Models , 2006, J. Mach. Learn. Res..

[9]  Olivier Ledoit,et al.  Improved estimation of the covariance matrix of stock returns with an application to portfolio selection , 2003 .

[10]  Pablo A. Parrilo,et al.  Latent variable graphical model selection via convex optimization , 2010, 2010 48th Annual Allerton Conference on Communication, Control, and Computing (Allerton).

[11]  Venkat Chandrasekaran,et al.  Gaussian Multiresolution Models: Exploiting Sparse Markov and Covariance Structure , 2010, IEEE Transactions on Signal Processing.

[12]  Bin Yu,et al.  High-dimensional covariance estimation by minimizing ℓ1-penalized log-determinant divergence , 2008, 0811.3628.

[13]  Roman Vershynin,et al.  Introduction to the non-asymptotic analysis of random matrices , 2010, Compressed Sensing.

[14]  Michael I. Jordan Graphical Models , 2003 .

[15]  Martin J. Wainwright,et al.  Noisy matrix decomposition via convex relaxation: Optimal rates in high dimensions , 2011, ICML.

[16]  Shiqian Ma,et al.  Alternating Direction Methods for Latent Variable Gaussian Graphical Model Selection , 2012, Neural Computation.

[17]  Adam J. Rothman,et al.  Sparse permutation invariant covariance estimation , 2008, 0801.4837.

[18]  Ambuj Tewari,et al.  Learning Exponential Families in High-Dimensions: Strong Convexity and Sparsity , 2009, AISTATS.

[19]  R. Tibshirani,et al.  Sparse inverse covariance estimation with the graphical lasso. , 2008, Biostatistics.

[20]  Alan S. Willsky,et al.  Learning Gaussian Graphical Models with Observed or Latent FVSs , 2013, NIPS.

[21]  Jonathan E. Taylor,et al.  On model selection consistency of M-estimators with geometrically decomposable penalties , 2013, NIPS 2013.

[22]  Yi Ma,et al.  Robust principal component analysis? , 2009, JACM.

[23]  Constantine Caramanis,et al.  Robust PCA via Outlier Pursuit , 2010, IEEE Transactions on Information Theory.

[24]  Vincent Y. F. Tan,et al.  Learning Latent Tree Graphical Models , 2010, J. Mach. Learn. Res..

[25]  P. Bickel,et al.  SIMULTANEOUS ANALYSIS OF LASSO AND DANTZIG SELECTOR , 2008, 0801.1095.