| || |
Submitted by pscully on Thu, 05/20/2010 - 16:57.
05/27/2010 - 16:10
05/27/2010 - 17:30
STA/BST 290: Nicolas Verzelen (INRA Montpellier)
Estimation of Gaussian graphs by model
Thursday, May 27th, 2010 at 4.10pm, MSB 1147 (Colloquium Room)
Refreshments: 3.30pm, MSB 4110 (Statistics Lounge)
Speaker: Nicolas Verzelen (INRA Montpellier, France, visiting UC Davis)
Title: Estimation of Gaussian graphs by model
Abstract: A current challenge in system biology is to infer the regulation network of a family of p genes from a n-sample of microarrays, with n (much) smaller than p. Gaussian graphical models are simple models to describe these regulation networks. If many works have been devoted to design fast estimation procedures of the graph, there is still much room for improvement: the existing procedures are only valid under restrictive assumptions. Moreover, the choice of their tuning parameters generally requires the knowledge of unknown quantities such as the conditional variance.
I will first describe the minimax rates of estimation for this problem. An interesting phase transition phenomenon occurs in the case of ultra-high dimensionality (p exponentially large with respect to n). Then, I will use model selection techniques as a way to tune and combine different graph estimation procedures.