PREDICT-EPFL/POP-BO

Implementation of the Principled Optimistic Preferential Bayesian Optimization (POP-BO) algorithm.

POP-BO

Code for the paper "Principled Preferential Bayesian Optimization" arXiv. POP-BO stands for 'Principled Optimistic Preferential Bayesian Optimization'.

Installation

Under this directory, run pip install ..

Getting started

You can run the demo.ipynb under "./popbo/demo"

Requirements

The required Python dependencies include:

• SafeOpt
• GPy

The following packages are also required:

• CasADi
• Ipopt

We also reused test functions in the repo:

• preferentialBO