Small Summaries, Efficient Algorithms and Fundamental Limits

Challenges abound in modern large scale data analysis, ranging from the sheer volume of the data to its complex structure and the need for timely responses. A promising common approach centers on capturing key information from the data using concise representations that can be constructed in a distributed manner or from an evolving stream of data. In this talk, we will illustrate both the power and limitations of this approach using three research vignettes. In the first example, we describe an algorithmic framework for fundamental linear algebra problems based on short summaries. In the second example, we design distributed algorithms for a family of non-convex problems arising in learning applications. In the third example, we show that even basic statistical estimation tasks require large summaries.
Huy Le Nguyen is a Research Assistant Professor at the Toyota Technological Institute at Chicago (TTIC). His research interests center on building theoretical foundation for processing modern large data sets. He has worked on dimensionality reduction, streaming algorithms, distributed algorithms, and their applications in machine learning. Huy received his BS in computer science and mathematics and MEng in computer science from MIT in 2008 and 2009, respectively, and his PhD in Computer Science from Princeton University in 2014 while being supported by a Gordon Wu Fellowship, a Siebel Scholarship, and an IBM PhD Fellowship. From 2014 to 2015, he was a Google Research Fellow at the Simons Institute at UC Berkeley.