README： Symbolic Regression via Control Variable Genetic Programming

This is the code implemetation of paper accepted at ECML-PKDD 2023 with title: Symbolic Regression via Control Variable Genetic Programming.

0. Prerequisites

0.1 Dependency packages

pip install deap

0.2 Directory

• data: the generated dataset. Every file represent a ground-truth expression.
• dso: public code implementation from https://github.com/brendenpetersen/deep-symbolic-optimization.
• plots: the jupter notebook to generate our figures in the paper.
• result: contains all the output of all the programs, the training logs.
• src: the inplemenattion of the our proposed control variable genetic programming algorithm and the classic genetic programming algorithm.

1. Run Control Variable Generic Programming (CVGP) and generic prgoramming (GP)

1.1 configure the basepath

Run the CVGP, GP model on the Noiseless [inv, sincos, sincosinv] dataset with configurations (5,5,8).

./src/scripts/run_gp_cvgp.sh

Assume we want to run CVGP, GP the Noisy [inv, sincos, sincosinv] dataset with configurations (5,5,8).

./src/scripts/noisy_run_gp_cvgp.sh

2. Run DSR, PQT, VPG, GPMeld

2.0 prequisites

1. install python environment 3.6.13: conda create -n py3613 python=3.6.13.
2. use the enviorment conda env py3613.
3. install dso

cd ./dso
pip install --upgrade setuptools pip
export CFLAGS="-I $(python -c "import numpy; print(numpy.get_include())") $CFLAGS"
pip install -e ./dso

3. create the .csv data file and .json model configuration file

If you

# generate the **Noiseless** **[inv, sincos, sincosinv]** dataset with configurations *(5,5,8)*.
./dso/dataset/gen_data.sh
# generate the **Noisy** **[inv, sincos, sincosinv]** dataset with configurations *(5,5,8)*.
./dso/dataset/noisy_gen_data.sh

4. run DSR, PQT, VPG, GPMeld models by If you want to run DSR, PQT, VPG, GPMeld on Noiseless datasets.

./dso/scripts/run_dsr_pqt_vpg_gpmeld.sh

If you want to run DSR, PQT, VPG, GPMeld on Noisy datasets.

./dso/scripts/noisy_run_dsr_pqt_vpg_gpmeld.sh

3. Look at the summarized result

Just open the plots folder

4. Parse your own results

./result/summary_output.sh

Cite

If you want to reuse this material, please considering citing the following:

 @InProceedings{10.1007/978-3-031-43421-1_11,
 author="Jiang, Nan
 and Xue, Yexiang",
 title="Symbolic Regression via Control Variable Genetic Programming",
 booktitle="Machine Learning and Knowledge Discovery in Databases: Research Track",
 year="2023",
 publisher="Springer Nature Switzerland",
 address="Cham",
 pages="178--195",
 isbn="978-3-031-43421-1"
 }