ECE Seminar: Leveraging inexactness for scalable optimization

Mankind generates data at a pace that vastly outstrips processing power, so much so that even "tractable" (polynomial time) algorithms are often too slow. Linear, or better yet, sublinear complexity is the most we can afford. These limitations induce a paradigm shift in how we approach modern data analysis: lower accuracy, randomization, online processing, distributed computation, inexact algorithms, and inexact models gain the centerstage.In my talk, I first review a few noteworthy examples from my own research that highlight the above approaches. To go somewhat deeper, I then describe more details of a versatile new, flexible, inexact optimization framework that I have recently developed.