mlr3mbo - Configuration

Code to configure and benchmark mlr3mbo using problems from YAHPO Gym.

Setup

• Run install.sh to setup the environment on the ncar derecho cluster
• Run install.R to install the R packages and python packages

Coordinate Descent

Files for running the coordinate descent are in the coordinate_descent folder.

• coordinate_descent_numeric.R run coordinate descent on the pure numeric subset of YAHPO Gym SO v1
• coordinate_descent_mixed.R run coordinate descent on the YAHPO Gym SO v1
• OptimizerCoordinateDescent.R bbotk optimizer for coordinate descent

The results of the coordinate descent are stored in coordinate_descent/results.

Benchmarking

Files for benchmarking mlr3mbo against SMAC3, Ax, HEBO and Optuna are in the competitors folder.

• run_numeric_mlr3mbo.R run mlr3mbo on the numeric benchmark instances
• run_numeric_competitors.R run SMAC3, Ax, HEBO and Optuna on the numeric benchmark instances
• run_mixed_mlr3mbo.R run mlr3mbo on the mixed benchmark instances
• run_mixed_competitors.R run SMAC3, Ax, HEBO and Optuna on the mixed benchmark instances
• wrapper_*.py python wrappers for SMAC3, Ax, HEBO and Optuna

The results of the benchmarking are stored in competitors/results.

Common files for the coordinate descent and benchmarking are in the common folder.