This repository was created for the conference paper "Enhancing the Computational Efficiency of Genetic Programming through Alternative Floating-Point Primitives," which was accepted to the 2024 Parallel Problem Solving from Nature (PPSN) conference.
Python-based implementations of the proposed math algorithms, along with the relevant experiments and results for measuring median relative error, are provided in code/approximations. Separately, a Jupyter notebook containing the results included for Operon within the paper is provided in code/operon. Lastly, files needed for locally running the Operon experiments and analysis notebook can be installed using the following instructions.
The following is only necessary when executing the relevant scripts locally. Note that the results included within the paper can be viewed without installation by simply opening the relevant CSV or Jupyter notebook files.
Also, when executing locally, note that only Linux operating systems are likely supported. The following was verified with insert CPU information and insert OS information.
- Ensure that some Conda package management system (e.g., Miniconda) is installed on the relevant machine.
- Download the latest software release from GitHub, available here. Extract and paste the contents of the
data.zipfile from the software release into thecode/operonfolder.
Upon doing the above, set up the relevant Conda environment and install the relevant tools by executing the following within a shell program, after having navigated to the repository directory within the shell:
conda env create -f environment.yml
conda activate conference-ppsn-2024
bash install.sh
In order to run the relevant Operon experiments, first navigate to the code/operon folder. Then, execute the following commands within your shell program:
for backend in {Eigen,Vdt,Stl,Mad_Transcendental_Fast,Mad_Transcendental_Faster,Mad_Transcendental_Fastest}; do ./operon_experiment.py --bin ./build_${backend}/cli/operon_nsgp --data ./problems/ --reps 100 >> my_results.csv; doneAfter the above finishes execution, insert details about running other relevant scripts.
Note: some of the results reported in the paper will definitely be irreproducible, e.g., runtime and energy measurements. In addition, if there are discrepancies in hardware/firmware, it is possible that there may exist other differences in results, e.g., due to differing implementations of the IEEE-754 floating-point standard.