Matrix algorithms: fast, stable, communication-optimizing--random?!

Ioana Dumitriu
University of Washington

What is the processor-memory gap, and why should one care that it increases? Can we find the complexity of matrix multiplication, and would it be of practical use to know it? How can randomization make an algorithm more reliable? These are all interesting and relevant questions in modern scientific computing, and I will try to address them from a broad perspective.

Presentation (PDF File)

Back to Workshop I: Convex Optimization and Algebraic Geometry