Skip to content
Bayesian optimization with the Gaussian process assumption
Branch: master
Clone or download

Latest commit

Fetching latest commit…
Cannot retrieve the latest commit at this time.


Type Name Latest commit message Commit time
Failed to load latest commit information.

GP-EST: A Unifying Algorithm for GP-UCB and Probability of Improvement

This is a matlab demonstration for an algorithm for Bayesian optimization with the Gaussian process assumption. The algorithm is fully described in

Optimization as Estimation with Gaussian Processes in Bandit Settings (Zi Wang, Bolei Zhou, Stefanie Jegelka), In International Conference on Artificial Intelligence and Statistics (AISTATS), 2016.

This paper is available at

To run the code, first install the gpml toolbox by Carl Rasmussen and Hannes Nickisch ( See gpo_example.m for a full explanation on how to run the algorithms.

In the paper, in addition to experiments on the synthetic functions, we also conducted experiments related to initialization tuning for trajectory optimization and parameter tuning for image classification. The trajectory optimization experiment is based on the Airplane2D example in the Drake toolbox at The image classification experiment follows the paper "Learning deep features for scene recognition using places database (Zhou et. al.)" in NIPS 2014. And all the datasets are available online.

You can’t perform that action at this time.