Bayesian Inference for Extended Latent Gaussian Models
Event details
| Date | 20.11.2020 |
| Hour | 16:15 › 18:00 |
| Speaker | Alex Stringer, University of Toronto |
| Location |
zoom
Online
|
| Category | Conferences - Seminars |
We define a novel class of additive regression models called Extended Latent Gaussian Models and derive inference methodology for this class which scales to large datasets and is accompanied by theoretical guarantees.
We demonstrate this approach on several examples including spatial downscaling, spatial proportional hazards regression with partial likelihood, and Gaussian Process classification, and demonstrate improved performance and scalability for existing Bayesian approaches to Generalized Linear Mixed and Additive Models.
A notable feature of the method is the natural manner in which we account for uncertainty in “smoothness” parameters, a well-known and difficult problem which we address in this talk through both application and theory.
Practical information
- Informed public
- Free
Organizer
- Prof. Anthony Davison
Contact
- Maroussia Schaffner