Skip to content

Toy code for reproducing experiments from the paper "Continuous Multi-Task Bayesian Optimization with Correlation"

License

Notifications You must be signed in to change notification settings

scrambledpie/Conditional-Multi-Task-BayesOpt

Repository files navigation

Continuous Multi-Task Bayesian Optimization with Correlation

Toy code for reproducing the simpler experiments from the paper:

Continuous Multi-Task Bayesian Optimization with Correlation

The code is in R and is all self-contained in a single file. It requires R packages which are installed by running these command in the R console:

install.packages("FastGP")
install.packages("MASS")
install.packages("Rcpp")

The algortihm takes a function of two inputs and finds all the optima of one input conditional on the other input. In other words, optimize every vertical slice in the picture below, each slice is it's own Bayesian optiimzation problem, we just assume there is more than one, or even a continuum of problems.

@article{pearce2018continuous,
  title={Continuous multi-task Bayesian Optimisation with correlation},
  author={Pearce, Michael and Branke, Juergen},
  journal={European Journal of Operational Research},
  volume={270},
  number={3},
  pages={1074--1085},
  year={2018},
  publisher={Elsevier}
}

alt text

alt text

About

Toy code for reproducing experiments from the paper "Continuous Multi-Task Bayesian Optimization with Correlation"

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Languages