Multi-Guiding Spark Fireworks Algorithm

NEWS

✨ We have released a GPU-accelerated version of MGFWA (https://arxiv.org/abs/2501.03944), available in alg/GPU-MGFWA.py. Please note that a custom neural network is required to fully utilize this optimization.

---

This repository is the official implementation of the paper Multi-Guiding Spark Fireworks Algorithm: Solving Multimodal Functions by Multiple Guiding Sparks in Fireworks Algorithm. The proposed method is powerful on multimodal global optimization, which is an efficient, performant, and state-of-the-art Fireworks Algorithm (FWA🎇) variant and easy to follow. Have fun enjoying it!

Installation 💡

Step1: clone this repository

git clone https://github.com/mxxxr/MGFWA.git

Step2: create a conda environment

conda create -n fwa python=3.8
conda activate fwa

Step3: install the dependencies

cd MGFWA
pip install numpy tqdm .

Run 🚀

Perform a stantard optimization process on the given benchmark (chosen from cec2013 and cec2017) by the given algorithm (chosen from MGFWA and LoTFWA):

python optimize.py --alg MGFWA --benchmark cec2013

The whole process will optimize each Function 51 times. After finishing the optimization of a single Funciton, the results will be output in the terminal:

Function #1, Optimizing...
MAX: 0.0
MIN: 0.0
MEAN: 0.0
MEDIAN: 0.0
STD: 0.0
Average runtime of a run: 17.03

Results 📈

Please refer to the paper for more quantitative results.

Data Profiles Method Analyzing

<img src="assets/DPM13_mean.png" width="33%" height="33%" /> <img src="assets/DPM13_std.png" width="33%" height="33%" />

<img src="assets/DPM17_mean.png" width="33%" height="33%" /> <img src="assets/DPM17_std.png" width="33%" height="33%" />

Contribution of the Multi-Guiding Spark

<img src="assets/contribution.png" width="50%" height="50%" />

Citation

If you found this repository useful, please consider citing our work:

@article{MENG2023101458,
  title = {Multi-guiding spark fireworks algorithm: Solving multimodal functions by multiple guiding sparks in fireworks algorithm},
  author = {Xiangrui Meng and Ying Tan},
  journal = {Swarm and Evolutionary Computation},
  pages = {101458},
  year = {2023},
  issn = {2210-6502},
  doi = {https://doi.org/10.1016/j.swevo.2023.101458},
  url = {https://www.sciencedirect.com/science/article/pii/S2210650223002304}
}

@misc{meng2025gpuimplementationmultiguidingspark,
      title={A GPU Implementation of Multi-Guiding Spark Fireworks Algorithm for Efficient Black-Box Neural Network Optimization}, 
      author={Xiangrui Meng and Ying Tan},
      year={2025},
      eprint={2501.03944},
      archivePrefix={arXiv},
      primaryClass={cs.NE},
      url={https://arxiv.org/abs/2501.03944}, 
}

License

This repository is licensed under Apache 2.0