I do not know what order these should be in or even if this is the entire interview. I did not know of either Margaret Wright or Michael Overton before stumbling on this. Apparently she was once president of the Society of Industrial and Applied Mathematicians, worked with George Dantzig at Stanford, and so on.
- Is linear programming polynomial time?
- Karmaker's algorithm
- Creating algorithms and writing software
- First job at GTE Sylvania
- Early years in California, Arizona and choosing Stanford
- Operations research at Stanford
- Faculty in math and computer science at Stanford
- Committees and SIAM
- Advisory Committees and the NAS
- Unexpected connections
- Accomplishements and progress for women
I think I have not appreciated how much progress has been made in the last few decades:
"... but look back and see the progress we've made. So in optimization at various points people have said well, for nonlinear problems, you can solve problems with hundreds of variables. Of course, that's a very imprecise statement. They never say what kind of problems or whatever. Today we had a talk where the person was talking about hundreds of thousands or millions of variables. That's an amazing change. It's an amazing change in capability. And it's not all due to computer hardware. I think sometimes people think, oh, machines are faster; machines are bigger. But it's smarter and better algorithms, which is the area you and I work in, right? Not making faster machines. We try to take advantage of them, but we don't make them run faster. And to me, that's astonishing. And the great part of it, I think there's still interesting problems that don't have many variables in this non-derivative optimization area." -- Margaret Wright
I do not understand interior point methods or why the Simplex algorithm works so well on average.
Posts
No comments:
Post a Comment