1. ELVES.py is the core implementation of our feature selection algorithm ELVES. It calls key functions from Functions.py. The GPU-accelerated version, ELVES_gpu.py, utilizes the GPU for computation and relies on Functions_gpu.py for core functions.
2. ManiFeSt.py is the main implementation of the algorithm ManiFeSt we use for comparison, and ManiFeSt_gpu.py is its GPU version.
3. The folder named XOR-100 contains the experimental code for reproducing the XOR-100 problem mentioned in our paper. Running XOR-100.py can get the experimental results of ELVES and ManiFeSt, and running XOR-100-baseline.py can get the experimental results of other baselines we compared.
4. The folder named Hypercube contains the experimental code for reproducing the Hypercube dataset mentioned in our paper. Running ELVES_Hypercube.py can get the experimental results of our algorithm ELVES, running ManiFeSt_Hypercube.py can get the experimental results of ManiFeSt, running ReliefF_IG_Hypercube.py can get the experimental results of the comparison methods ReliefF and IG, and running baseline_Hypercube.py can get the experimental results of other baselines we compared.
5. The folder named Prostate contains the experimental code for reproducing the Prostate dataset mentioned in our paper. Running ELVES_Prostate_gpu.py can get the experimental results of our algorithm ELVES, running baseline(tuning)_prostate_gpu.py can get the experimental results of the comparison methods ManiFeSt, ReliefF, IG and Laplacian Score, and running baseline_prostate_gpu.py can get the experimental results of other baselines we compared.
6. It should be noted that the baseline method we compared is mainly implemented using functions in the skfeature library. The skfeature library needs to be installed before running the code file. The specific experimental settings are shown in Appendix A of our paper.