|
Activity Number:
|
134
- Bayesian Modeling
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 9, 2021 : 1:30 PM to 3:20 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
|
Abstract #318496
|
|
|
Title:
|
Scalable Bayesian Optimization Using Ordered Conditional Approximations of Gaussian Processes
|
|
Author(s):
|
Felix Jimenez* and Matthias Katzfuss
|
|
Companies:
|
Texas A&M University and Texas A&M University
|
|
Keywords:
|
reinforcement learning;
Bayesian Optimization;
Gaussian processes;
Vecchia approximation;
Sparse Inverse Cholesky factor
|
|
Abstract:
|
Bayesian optimization is a technique for optimizing black-box target functions. At the core of Bayesian optimization is a surrogate model that predicts the output of the target function at a previously unseen input to facilitate the selection of promising input values. Gaussian processes (GPs) are a common surrogate model but are known to scale poorly with the number of observations. To address this scaling issue, we propose the use of an ordered conditional GP approximation. Our approximation is well suited for extensions such as the selection of multiple input values in parallel. We showcase the advantages of our approximation relative to existing methods in several numerical comparisons. We end with a discussion on Bayesian optimization in higher dimensions.
|
Authors who are presenting talks have a * after their name.
