Feizi Wins Best Paper Award from IEEE Transactions on Network Science and Engineering

Wed Apr 24, 2019

Soheil Feizi, an assistant professor of computer science with an appointment in the University of Maryland Institute for Advanced Computer Studies, has received a Best Paper Award from the journal IEEE Transactions on Network Science and Engineering.

“Network Maximal Correlation,” published in 2017 by Feizi and researchers from MIT and the National University of Ireland Maynooth, provides a natural extension of the maximal correlation concept to more than two variables, as is often the case in modern big data applications.

It introduces a new statistic for measuring dependency in multivariate random variables as well as providing a computational framework to evaluate it given multivariate data sets.

Identifying relationships among variables in large datasets is an increasingly important task in systems biology, social sciences, finance, and other fields, Feizi says.

The most well-known association measure between two variables is Pearson’s correlation, which is based on a linear fit of data. A nonlinear generalization of Pearson's correlation is “maximal correlation,” invented in the 1950s to overcome the linearity drawbacks of Pearson’s correlation.

The goal of this research is to automatically identify complex, non-linear dependencies that would not be observed with traditional measures of correlation.