Neural Exploratory Landscape Analysis

Requirements

You can install all of dependencies of NeurELA via the command below.

pip install -r requirements.txt

After installing these dependecies, please refer to instruction_to_modify_pypop7 for help to modify the source code of PyPop7 to meet our requirements.

Train

The training process can be easily activated via the command below.

python run.py

For more adjustable settings, please refer to run.py and config.py for details.

Recording results: Log files and NeurELA model checkpoints will be saved to ./records, the file structure is as follow.

records
|--run_name
   |--log_file
      |--logging files
      |--...
   |--saved_model
      |--model checkpoint
      |--...

Zero-shot

Once the NeurELA checkpoint saved, you can validate its zero-shot performance via running the commmand below. Note that you should provide load_path correctly

python zero-shot.py

You can modify testsuits or MetaBBO algorithms in zero-shot.py for specific requirements.

Fine-tune

You can activate the fine-tuning process by running the command below,

python transfer.py

Similar to zero-shot process, you can modify testsuits or MetaBBO algorithms in transfer.py for specific requirements.

Citing NeurELA

The PDF version of the paper is available here. If you find our NeurELA useful, please cite it in your publications or projects.

@article{ma2024neural,
  title={Neural Exploratory Landscape Analysis},
  author={Ma, Zeyuan and Chen, Jiacheng and Guo, Hongshu and Gong, Yue-Jiao},
  journal={arXiv preprint arXiv:2408.10672},
  year={2024}
}