Farinaz Koushanfar

ECE Assistant Professor
Rice University

Hierarchical Non-Parametric Approach for Recovery for Missing Sensor Data

Recovery of missing data is a canonical task in sensor networks, wireless communications, design space exploration, active learning, and many other research and engineering areas. At best, many important tasks in sensor networks such as fault detection, data cleaning, sampling, compression, and topology management are addressed by utilizing missing data recovery methods. Numerous techniques have been proposed for missing data recovery, including nonlinear function minimization, clustering, maximum likelihood, Monte Carlo simulation techniques, and variational methods. However, most of the proposed techniques place strong assumptions on the sensor data and its distributions. Often, such assumptions about the data are not justified for the recovery of missing sensor data and are rarely effective.

This talk unveils a conceptually new approach for the recovery of missing sensor measurements that imposes a very mild set of assumptions on the sensor data. The procedure hierarchically applies the standard regression and classification techniques combined within a framework that enables rapid statistical learning and validation. The approach consistently follows a paradigm of using nonparametric statistical techniques based on data-driven models in all steps of the missing data recovery. The effectiveness of the technique is demonstrated for data recovery and other computational sensing tasks on the traces of actually deployed sensor networks.
Tuesday, September 20, 2005
10:15a.m. - Duncan Hall, McMurtry Auditorium
Rice University
* Biography:

Farinaz Koushanfar received her BSEE in 1998 from Sharif University of Technology, MSEE in 2000 from the University of California, MA in Statistics, University of California, in 2004 and PhD in Electrical Engineering and Computer Science from the University of California, Berkeley in 2005. Her research interests include sensor networks, wireless embedded systems, as well as optimization and statistics.

ECE Affiliates Meeting - Morning Session



Last modified: September 26, 2005