Friday, 15:15 - 15:40 h, Room: MA 376

 

Stefanie Jegelka
On fast approximate submodular minimization and related problems

Coauthors: Jeff A. Bilmes, Hui Lin

 

Abstract:
Machine learning problems often involve very large data sets. To test algorithms quickly, we aim to extract a suitable subset of a large training corpus. This is a submodular minimization problem, but the size of the data renders current exact methods very impractical. Graph cuts can be an alternative, but may not be able to efficiently represent any submodular function. We therefore approximate the objective function by a sequence of graph-representable functions. This leads to an efficient approximate minimization algorithm. It turns out that the underlying model not only helps represent submodular functions, it also enhances applications of graph cuts in computer vision, representing non-submodular energy functions that improve image segmentation results.

 

Talk 1 of the invited session Fri.3.MA 376
"Methods from discrete mathematics in systems biology" [...]
Cluster 12
"Life sciences & healthcare" [...]

 

  Payday Loans Florida can help you in trying times, but be sure to know the laws necessary for your loan application. But it is worth noting that these tests were carried out on the blood cells. Therefore, it's too early to say about scientific evidence of Viagra influence on blood clots.