Laura Balzano receives AFOSR Young Investigator Award for research that addresses massive streaming data

Prof. Balzano uses statistical signal processing, matrix factorization, and optimization to unravel dynamic and messy data.

Laura Balzano Enlarge

Prof. Laura Balzano received a Young Investigator Award from the Air Force Office of Scientific Research (AFOSR) to support research modeling dynamic massive data. Her project is called “Non-convex Optimization Algorithms and Theory for Matrix Factorization with Dynamic Massive Data.”

This research addresses a problem that has eluded solutions using machine learning techniques, which do not take into account changing systems, instead assuming that the underlying models are static. For example, says Balzano, face recognition software assumes a single model of your face regardless of age; online marketing assumes older purchases are just as informative as newer ones; and social networks predict future interactions based on one’s sometimes distant past history.

“In reality,” explains Balzano, “all of these tasks are time-varying and dynamic. Most algorithms deal with this by simply looking at an abbreviated version of history, but there is little understanding of how to model these phenomena explicitly as dynamic phenomena, and what window of historical data is useful for inferring current reality.”

A key tool for modeling dynamic massive data has been the use of time-varying low-rank factorization models, typically using non-convex formulations. However, while this method is fast and can be applied to real-time dynamic problems, it offers few guarantees of correctness and convergence.

Balzano plans to investigate a broad range of non-convex formulations for streaming low-rank matrix factorization on dynamic, massive data and create classes of algorithms that can solve them quickly, accurately, and reliably with theoretical guarantees. These new algorithms, able to handle dynamic streaming data, missing data, and robust regularization, will revolutionize practice in a great many fields, especially those collecting massive streaming data.

Balzano’s research projects are in statistical signal processing, matrix factorization, and optimization, particularly dealing with large and messy data. She directs the Signal Processing Algorithm Design and Analysis (SPADA) lab, which studies algorithms for statistical signal processing and machine learning with applications in data analysis, computer vision, environmental monitoring, image processing, control systems, power grids, genetic expression data analysis, consumer preference modeling, and computer network analysis.

Balzano has also received an Intel Early Career Faculty Honor Program Award and a 3M Non-Tenured Faculty Award. She joined the University of Michigan in 2013 after receiving her doctoral degree from the University of Wisconsin, Madison.


AFOSR Press Release

The Air Force Young Investigator Program, initiated in 2006, supports scientists and engineers in the early stage of their careers who show exceptional ability and promise for conducting basic research.

This year, 31 scientists and engineers from 24 research institutions will receive $13.9M in grants. More than 290 proposals were submitted. Balzano is one of four University of Michigan researchers to receive the award.