Material Detail
What Do Unique Games, Structural Biology and the Low-Rank Matrix Completion Problem Have In Common
This video was recorded at Machine Learning Summer School (MLSS), Chicago 2009. We will formulate several data-driven applications as MAX2LIN and d-to-1 games, and show how to (approximately) solve them using efficient spectral and semidefinite program relaxations. The relaxations perform incredibly well in the presence of a large number of outlier measurements that cannot be satisfied. We use random matrix theory to prove that the algorithms almost achieve the information theoretic Shannon bound. The underlying group structure of the different applications (like SO(2), SO(3), GL(n), etc.) is heavily exploited. Applications include: cryo-electron microscopy and NMR spectroscopy for 3D protein structuring, low-rank matrix completion, clock synchronization, and surface reconstruction in computer vision and optics. Partly joint with Yoel Shkolnisky, Ronald Coifman and Fred Sigworth (Yale); Mihai Cucuringu and Yaron Lipman (Princeton); and Yosi Keller (Bar Ilan).
Quality
- User Rating
- Comments
- Learning Exercises
- Bookmark Collections
- Course ePortfolios
- Accessibility Info
More about this material
- Mobile Compatibility:
- Not specified at this time
- Language:
- English
- Cost Involved:
- No
- Source Code Available:
- No
- Creative Commons:
-
This work is licensed under a Attribution-NonCommercial-NoDerivs 3.0 United States
