EE Systems Seminar

The current data-age is witnessing an unprecedented confluence of

disciplines, blurring traditional domain boundaries. But what

aspects of data are driving this rich interaction? We can single out

at least two: size and structure.

Today, I will talk about both size and structure of data. In

particular, I will demonstrate how a large number of machine

learning problems (for instance regularized risk minimization,

dictionary learning and matrix factorization) fall into a generic

framework for scalable nonconvex optimization. I will highlight a

few applications that have benefited from this framework, while

commenting on ongoing and future work that strives for even greater

scalability.

Beyond size, I shall talk about "structure", specifically geometric

structure of data. My motivation lies in a number of applications of

machine learning and statistics to data that are not just vectors,

but richer objects such as matrices, strings, functions, graphs,

trees, etc. Processing such data in their "intrinsic representation"

can be of great value. Notably, we'll see examples where exploiting

the data geometry allows us to efficiently minimize a class of

nonconvex cost functions, not to local, but to global optimality.

Time permitting, I will also mention some surprising connections of

our work to areas beyond machine learning and data analysis.

Bio

Suvrit Sra is a Sr. Research Scientist at the Max Planck Institute

for Intelligent Systems, in Tübingen, Germany. He obtained his

Ph.D. in Computer Science from the University of Texas at Austin. He

has held visiting faculty positions at Carnegie Mellon University

(2013-14) in the Machine Learning Department and at UC Berkeley

(2013) in EECS.

His research involves several data-driven real-world applications,

which rely on a variety of tools from different mathematical areas

such as geometry, analysis, statistics, and optimization.

His work has won several awards; notably, the "SIAM 2011 Outstanding

Paper Prize". He regularly organizes a workshops on "Optimization

for Machine Learning" at the well-known Neural Information

Processing Systems (NIPS) conference, and has recently (co)-edited a

book with the same title.