Group:Student Seminar
Title: Matrix Multiplication
Speaker: Zihe Zhang University
Time: 2011-12-27 16:00-2011-12-27 17:00
Venue: FIT 1-222


Recently I'm reading materials on matrix multiplication. People prove that if we can calculate matrix multiplication faster, then we can calculate matrix inversion and the determinant faster. Then there comes the question:how many multiplication do we need to calculate two N*N matrix multilication? N^3 is of course enough using the basic method. In 1969, Strassen showed that n^{log)_2(7)} is enough. In 1987 Coppersmith -Winograd showed that n^{2.376} is enough. This year Virginia Vassilevska Willians showed that n^{2.3727} is enough in "Breaking the Coppersmith-Winograd barrier". Matrix multiplication is a so widely used algorithm in practice, it's very amazing that it even only improves from 2.376 to 2.373.

It's hard for me to really understand it, and there is still points I don't know. But I can explain how they get n^{2.376} and the way they use to get 2.3727.I will show it to you. I will try my best and not waste your time. 


Short Bio: