| Title |
Noisy global Bayesian optimization using generalized product of experts / |
| Authors |
Tautvaišas, Saulius ; Žilinskas, Julius |
| Full Text |
|
| Is Part of |
Proceedings of the Hungarian global optimization workshop HUGO 2022 / edited by B. G.-Tóth, T. Csendes.. Szeged : University of Szeged. 2022, p. 185-188 |
| Keywords [eng] |
Bayesian optimization ; generalized product of experts ; noise |
| Abstract [eng] |
In many real world optimization problems observations are very noisy, which require efficient optimization methods that can successfully handle varying levels of noise. Bayesian optimization (BO) is an efficient approach for global optimization of black-box functions, but the performance of using a Gaussian process (GP) model can degrade with changing levels of noise due homoscedastic noise assumption in the GP prior. However, generalized product of experts (gPoE) model build independent GP experts on the subsets of observations with individual set of hyperparameters, which is flexible enough to capture the changing levels of noise. We compare and evaluate the performance of gPoE based BO (gPoEBO) model on 6 synthetic global optimization functions with varying levels of homoscedastic and heteroscedastic noise. The results show that gPoEBO is competitive with other models and is able to improve the performance on functions with high homoscedastic noise. |
| Published |
Szeged : University of Szeged |
| Type |
Conference paper |
| Language |
English |
| Publication date |
2022 |
