PSGO-Optimizer

PSGO: A Novel Asymmetric Dual-Agent Swarm Optimization Inspired by Pistol Shrimp–Goby Fish Mutualism for Feature Selection and Engineering Design Optimization

Official implementation of PSGO (Pistol Shrimp–Goby Fish Optimization), a biologically inspired asymmetric dual-agent swarm optimization algorithm designed for global optimization, feature selection, and engineering design applications.

---

Authors

• Nishi Madan
• Rahul Malik

---

Abstract

PSGO is a novel asymmetric dual-agent swarm optimization algorithm inspired by the mutualistic relationship between pistol shrimp and goby fish. The proposed optimizer integrates biologically grounded asymmetric cooperation, a graded three-level danger signaling mechanism, and a claw-blast perturbation strategy to balance exploration and exploitation effectively. PSGO demonstrates strong optimization performance on benchmark functions, feature selection tasks, and engineering design optimization problems.

---

Key Contributions

• Biologically grounded asymmetric dual-agent swarm architecture.
• Novel graded three-level danger signal mechanism.
• Claw-blast perturbation strategy with adaptive decay radius.
• Strong performance on CEC benchmark functions.
• Applications to feature selection and engineering design optimization.

---

Repository Structure

PSGO-Optimizer/
│
├── src/
├── notebooks/
├── experiments/
├── results/
├── figures/
├── paper/
├── docs/
├── data/
│
├── requirements.txt
├── CITATION.cff
├── LICENSE
└── README.md

---

Installation

git clone https://github.com/nishimaliknitj/PSGO-Optimizer.git

cd PSGO-Optimizer

pip install -r requirements.txt

---

Usage

Run the optimizer:

python psgo.py

Run benchmark experiments:

python run_cec2017.py

Execute complete experiments:

python run_all_fast.py

---

Benchmark Evaluation

The proposed PSGO algorithm is evaluated on standard numerical optimization benchmarks including:

• CEC 2017 Benchmark Suite
• Feature Selection Datasets
• Engineering Design Optimization Problems

---

Results

This directory contains all experimental outputs generated by PSGO.

Files

cec2017_results.csv

Raw benchmark results for all algorithms on CEC 2017 functions.

Columns:

• Function
• Algorithm
• Run
• Error

friedman_ranking.csv

Average Friedman ranks of all algorithms.

Columns:

• Algorithm
• Friedman_Rank

wilcoxon_tests.csv

Wilcoxon signed-rank comparison between PSGO and competitors.

Columns:

• Competitor
• Wins
• Ties
• Losses

feature_selection.json

Feature selection experimental results.

engineering.json

Engineering design optimization results.

Engineering Design Problems

The repository contains implementations and experiments for engineering optimization tasks reported in the paper.

---

Reproducibility

All experiments, notebooks, benchmark scripts, and result generation codes required to reproduce the reported findings are included in this repository.

---

Citation

If you use PSGO in your research, please cite the associated paper.

@article{PSGO2026,
  title={PSGO: A Novel Asymmetric Dual-Agent Swarm Optimization Inspired by Pistol Shrimp--Goby Fish Mutualism for Feature Selection and Engineering Design Optimization},
  author={Madan, Nishi and Malik, Rahul},
  year={2026}
}

---

Contact

Rahul Malik

Email: maliknit@gmail.com