EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
427. The exponential cone: machine learning and the projection algorithm
Invited abstract in session MD-38: Applications of conic optimization, stream Conic Optimization: Theory, Algorithms, and Applications.
Monday, 14:30-16:00Room: 34 (building: 306)
Authors (first author is the speaker)
| 1. | Henrik A. Friberg
|
| MOSEK ApS |
Abstract
The exponential cone is an absolute beast with respect to its usages in mathematical modeling, and has proven to be a computational stable and high-performant element of the MOSEK optimization core. In this talk we will explore some of its surprising use cases, e.g. in machine learning, before unveiling the successful story of the "projection onto the exponential cone" algorithm. The Moreau decomposition theorem, the KKT conditions, the convex analysis book by Rockafellar, unconstrained root finding and floating-point safeguards are all parts of this story.
Keywords
- Mathematical Programming
- Convex Optimization
- Computer Science/Applications
Status: accepted
Back to the list of papers