Regression via Iterated Local Search for Symbolic Regression GECCO Competition -- 2023
Project description
RILS -- Regression via Iterated Local Search
RILS algorithm for GECCO2023 SR competition
Installation instructions
-
The program requires Python 3 (tested on version 3.11.3), but it should work on some earlier versions as well. We also recommend using pip package manager.
-
Download the repository and unpack it.
-
Install the following pip packages:
pip install numpy
pip install sympy
pip install scikit-learn
pip install statsmodels
Execution instructions
-
Position inside the root directory of the unpacked repository. The (sub)directories at this level are: instances/, paper/, results/ and rils/. Directory instances/ holds datasets that are previously rounded to 9 decimals and saved in the tab separated .txt files.
-
Call run_all.cmd or run_all.sh script depending on whether you use Windows or Linux. If using Linux, you might need to change the accesss rights for this script, i.e. add x (execution) rights. Also, you might need to change the name of python executable to python3 on some systems. The content of run_all.{cmd|sh} file is as follows:
python run.py "instances/srbench_2023" "dataset_1.txt" 180 20
python run.py "instances/srbench_2023" "dataset_2.txt" 100 10
python run.py "instances/srbench_2023" "dataset_3.txt" 180 20
-
The steps in the execution pipeline are explained in paper/rils_gecco2023.pdf. Briefly, several output files will occurr during execution:
- best_sols_dataset_{1|2|3}.txt
- log.txt
- out.txt
The model will be written inside out.txt. Note that there is semi-automated (interactive) simplification (step 4 in the paper) for the dataset_3.
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.