Stochastic Gradient Descent for Kernel-Based Maximum Correntropy Criterion
Tiankai Li, Baobin Wang, Chaoquan Peng, Hong Yin, Martin Trapp, Pierre Alquier · 2024
Read Stochastic Gradient Descent for Kernel-Based Maximum Correntropy Criterion by Tiankai Li, Baobin Wang, Chaoquan Peng, Hong Yin, Martin Trapp, Pierre Alquier (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-11675914.txt.
Description
Tiankai Li and co-authors analyzed a learning algorithm for a robust training criterion in a paper published in two thousand twenty-four. The criterion, maximum correntropy, uses higher-order information and resists outliers better than least squares. Its kernel form suits nonlinear models, but the optimization is nonconvex.
They study stochastic gradient descent, the workhorse method for such problems, applied to the kernel criterion. Past theory for this pairing was incomplete. Under standard conditions, they prove how the iterates behave and give explicit convergence rates toward the best solution. The work closes a gap between the optimization steps a learner takes and the guarantees that its estimate is truly optimal.
Catalog record
| CREATOR | Tiankai Li, Baobin Wang, Chaoquan Peng, Hong Yin, Martin Trapp, Pierre Alquier |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 21 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |