April 14 Meeting of the San Francisco Bay Area Chapter of the American Statistical Association

Speaker: Art Owen, Professor of Statistics, Stanford University, Joint work with Ya Xu and Justin Dyer, Stanford

Title: Empirical stationary correlations for semi-supervised learning on graphs

Time: Wednesday, April 14,  4 - 6 PM light refreshments 4 - 4:30, presentation 4:30 – 6 PM, followed by dinner hosted by Google, 6:30 PM.

Location: Google, 1950 Alza, Mountain View, CA

Abstract and Directions:

Abstract: In semi-supervised learning on graphs, response variables observed at one node are used to estimate missing values at other nodes. The methods exploit correlations between nearby nodes in the graph.  We prove that many such proposals are equivalent to kriging predictors based on a covariance matrix driven by the link structure of the graph.  We then propose a data-driven estimator of the correlation structure that exploits patterns among the observed response values.  By incorporating even a small fraction of observed covariation into the predictions we are able to obtain much improved prediction on two graph datasets.

Directions: 1950 Alza is the building right next door to the main Google campus at 1600 Amphitheatre Parkway in Mountain View.  Guests can use the parking valet and then walk over the bridge to 1950.  Please see this map.

Return to Bay Area ASA Homepage