Code to reproduce the experiments in the paper Semantic variation operators for multidimensional genetic programming
Experiments can be run using analysis/ml-analyst/submit_jobs.py
. See the command line options (python submit_jobs.py -h
) for help.
As example, this command would launch the entire experiment:
python submit_jobs.py --r -ml FeatTuned,FeatSXOTuned,FeatRXOTuned,MLPmod,Linear,XGBoost -n_trials 10 -data ../penn-ml-benchmark/datasets/ -results results/
analysis/
contains these notebooks:
results_tuning.ipynb
produces the tuning results figures.results_benchmark.ipynb
produces the comparisons to the Where are we now? paper.stats.ipynb
depends onresults_benchmark.ipynb
and `results_tuning.ipynb', produces the statistical tests.results_benchmark-extended.ipynb
contains code to reproduce the extended PMLB results.
- scikit-learn
- xgboost
- datasets come from PMLB
- FEAT