2009 Research Summary

Sparse Principal Component Analysis (SPCA)

View Current Project Information

Laurent El Ghaoui and Michael Jordan

In this work, we seek to decompose a matrix into sparse principal components. This has applications, for example, in genomics, allowing us to identify a few genes that explain most of the variance observed in micro-array data. We investigate various algorithms based on convex relaxations, their complexity, and the quality of the corresponding approximations.