Gp Optimization
The python code provided here includes several optimization algorithms (purely sequential or batch) using Gaussian processes. Available algorithms include GP-UCB, EI, Chaining-UCB for sequential optimization, and GP-UCB-PE, GP-B-UCB for batch optimization. The implementation uses efficient updating formulae in order to remain scalable to a large number of training points.
Documentation
Please go to http://econtal.perso.math.cnrs.fr/software/ for example of use and documentation.
Relevant publications
Statistical Learning Approaches for Global Optimization
Emile Contal
PhD
http://econtal.perso.math.cnrs.fr/publications/phd.pdf
Stochastic Process Bandits: Upper Confidence Bounds Algorithms via Generic Chaining
Emile Contal, Nicolas Vayatis
http://arxiv.org/pdf/1602.04976v1.pdf
A Ranking Approach to Global Optimization
Cédric Malherbe, Emile Contal, Nicolas Vayatis
ICML 2016
http://arxiv.org/pdf/1603.04381v1.pdf
Optimization for Gaussian Processes via Chaining
Emile Contal, Cédric Malherbe, Nicolas Vayatis
NIPS Workshop on Bayesian Optimization
http://arxiv.org/pdf/1510.05576v1.pdf
Gaussian Process Optimization with Mutual Information
Emile Contal, Vianney Perchet, Nicolas Vayatis
ICML 2014
http://jmlr.csail.mit.edu/proceedings/papers/v32/contal14.pdf
Parallel Gaussian Process Optimization with Upper Confidence Bound and Pure Exploration
ECML 2013
Emile Contal, David Buffoni, Alexandre Robicquet, Nicolas Vayatis
http://arxiv.org/pdf/1304.5350v3.pdf