EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
2954. Bilevel learning optimization and applications
Invited abstract in session TB-32: Nonsmooth optimization and applications, Part II, stream Advances in large scale nonlinear optimization.
Tuesday, 10:30-12:00Room: 41 (building: 303A)
Authors (first author is the speaker)
| 1. | Serena Crisci
|
| Department of Mathematics and Physics, University of Campania "L. Vanvitelli" |
Abstract
Bilevel optimization has gained a lot of attention in recent years as a data-driven learning technique for the estimation of unknown model operators or for the automatic selection of hyperparameters in highly parameterized models. Formally, a bilevel problem consists in a nested optimization problem where a variational model acts as a constraint.
Starting from an outline of the main theoretical concepts of bilevel formulations, in this talk numerical bilevel schemes to address imaging and portfolio optimization problems will be anaylsed, combined with different strategies to properly derive the gradient of the upper-level loss, which is one of the major obstacle in bilevel optimization. Numerical tests will be presented to validate the proposed schemes.
Keywords
- Convex Optimization
- Non-smooth Optimization
Status: accepted
Back to the list of papers