B0435
Title: The functional graphical lasso
Authors: Kartik Waghmare - EPFL (Switzerland)
Tomas Masak - EPFL (Switzerland) [presenting]
Victor Panaretos - EPFL (Switzerland)
Abstract: The problem of recovering conditional independence relationships between p jointly distributed Hilbertian random elements is considered given n realizations thereof. It is operated in the sparse high-dimensional regime, where $n << p$ and no element is related to more than $d << p$ other elements. In this context, an infinite-dimensional generalization of the graphical lasso is proposed. Model selection consistency is proven under natural assumptions and many classical results to infinite dimensions are extended. In particular, finite truncation or additional structural restrictions are not required. The plug-in nature of the method makes it applicable to any observational regime, whether sparse or dense, and indifferent to serial dependence. Importantly, the method can be understood as naturally arising from a coherent maximum likelihood philosophy.
