Eric Hans Lee
Research Subjects
Eric Hans Lee has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
- Aleksei G. Sorokin
Illinois Institute of Technology, USA
1 shared publication Bolong Cheng
1 shared publicationCameron Ibrahim
1 shared publication- David Bindel
Cornell University, Ithaca, USA
1 shared publication Leo Huang
1 shared publication- Xinran Zhu1 shared publication
Latest Publications
- 29 April 2023
- Published by Elsevier in Knowledge-Based Systems
- Vol. 273
SigOpt Mulch: An intelligent system for AutoML of gradient boosted trees
- 29 April 2023
- Published by Elsevier in Knowledge-Based Systems
- Vol. 273
SigOpt Mulch: An intelligent system for AutoML of gradient boosted trees
- Open Access
Scalable Bayesian Transformed Gaussian Processes
The Bayesian transformed Gaussian process (BTG) model, proposed by Kedem and Oliviera, is a fully Bayesian counterpart to the warped Gaussian process (WGP) and marginalizes out a joint prior over input warping and kernel hyperparameters. This fully Bayesian treatment of hyperparameters often provides more accurate regression estimates and superior uncertainty propagation, but is prohibitively expensive. The BTG posterior predictive distribution, itself estimated through high-dimensional integration, must be inverted in order to perform model prediction. To make the Bayesian approach practical and comparable in speed to maximum-likelihood estimation (MLE), we propose principled and fast techniques for computing with BTG. Our framework uses doubly sparse quadrature rules, tight quantile bounds, and rank-one matrix algebra to enable both fast model prediction and model selection. These scalable methods allow us to regress over higher-dimensional datasets and apply BTG with layered transformations that greatly improve its expressibility. We demonstrate that BTG achieves superior empirical performance over MLE-based models.
Open AccessScalable Bayesian Transformed Gaussian Processes
The Bayesian transformed Gaussian process (BTG) model, proposed by Kedem and Oliviera, is a fully Bayesian counterpart to the warped Gaussian process (WGP) and marginalizes out a joint prior over input warping and kernel hyperparameters. This fully Bayesian treatment of hyperparameters often provides more accurate regression estimates and superior uncertainty propagation, but is prohibitively expensive. The BTG posterior predictive distribution, itself estimated through high-dimensional integration, must be inverted in order to perform model prediction. To make the Bayesian approach practical and comparable in speed to maximum-likelihood estimation (MLE), we propose principled and fast techniques for computing with BTG. Our framework uses doubly sparse quadrature rules, tight quantile bounds, and rank-one matrix algebra to enable both fast model prediction and model selection. These scalable methods allow us to regress over higher-dimensional datasets and apply BTG with layered transformations that greatly improve its expressibility. We demonstrate that BTG achieves superior empirical performance over MLE-based models.