Research Buffet
Friday, September 2, 2011
3147 MEB
Refreshments 3:40 p.m.
Lecture 4:00 p.m.
Suresh Venkatasubramanian
Assistant Professor
Title: From pattern matching to high dimensional geometry and back
Friday, September 2, 2011
3147 MEB
Refreshments 3:40 p.m.
Lecture 4:00 p.m.
Suresh Venkatasubramanian
Assistant Professor
Title: From pattern matching to high dimensional geometry and back
Abstract
Data mining is about finding patterns in data. It turns out that these patterns "live" in a very high dimensional space. To find patterns, we therefore need to understand "shapes" in high dimensional space. I'll talk about the geometric perspective on data mining, and the strange non-Euclidean geometries you sometimes find yourself in.
Matthew Might
Assistant Professor
Title: A brief introduction to static analysis
Data mining is about finding patterns in data. It turns out that these patterns "live" in a very high dimensional space. To find patterns, we therefore need to understand "shapes" in high dimensional space. I'll talk about the geometric perspective on data mining, and the strange non-Euclidean geometries you sometimes find yourself in.
Matthew Might
Assistant Professor
Title: A brief introduction to static analysis
Abstract
Through the Halting Problem, the very inception of computer science is intimately tied to what computers cannot do. In short, the Halting Problem forbids us from writing programs that precisely predict the behavior of other programs. This is unfortunate, for if this were possible, many problems in software reliability, security and optimization would vanish.
The entire enterprise of static analysis is a quixotic venture to show that the Halting Problem holds only in the limit--that for programs of human interest, purely mechanical reasoning can efficiently predict and bound their behavior.
Come learn how it's done, how far we've come and where the frontier in automated reasoning about programs lies.
Through the Halting Problem, the very inception of computer science is intimately tied to what computers cannot do. In short, the Halting Problem forbids us from writing programs that precisely predict the behavior of other programs. This is unfortunate, for if this were possible, many problems in software reliability, security and optimization would vanish.
The entire enterprise of static analysis is a quixotic venture to show that the Halting Problem holds only in the limit--that for programs of human interest, purely mechanical reasoning can efficiently predict and bound their behavior.
Come learn how it's done, how far we've come and where the frontier in automated reasoning about programs lies.