244. Computing Augustin Information via Hybrid Geodesically Convex Optimization
Invited abstract in session WE-4: Large-scale optimization I, stream Large-scale optimization.
Wednesday, 14:10 - 15:50Room: M:M
Authors (first author is the speaker)
| 1. | Guan-Ren Wang
|
| National Taiwan University |
Abstract
We propose a Riemannian gradient descent with the Poincaré metric to compute the Augustin information, a widely used quantity for characterizing exponential error behaviors in information theory. We provide a non-asymptotic optimization error guarantee for this algorithm. Our result is based on a novel hybrid analysis of Riemannian gradient descent for functions that are geodesically convex in a Riemannian metric and geodesically smooth in another. Numerical experimental results demonstrate the empirical efficiency of the algorithm.
Keywords
- Large- and Huge-scale optimization
- Global optimization
Status: accepted
Back to the list of papers