Kyushu University Institute of Mathematics for Industry

To analyze large-scale data such as gene expression data, we often need a statistical model which consists of a large number of parameters (e.g., hundreds of millions.) The sparse estimation, such as the lasso, makes most of the parameters exactly zero, enabling an efficient extraction of useful information from the data. Recently, I am interested in the development of new sparse estimation procedures in multivariate analysis, such as the factor analysis and the Gaussian graphical modeling. Specifically, I am developing several numerical algorithms that efficiently compute the estimate of the parameter, and also investigating theoretical properties of the estimated parameters. Most of the proposed methods are available for use in the R packages.

Keywords Sparse Estimation, Multivariate Analysis
Division Division of Industrial and Mathematical Statistics, Division of Strategic Liaison (Concurrent), Division of Fujitsu Mathematical Modeling for Decision Making (Concurrent)