theoverhelst/aco

Project on Ant Colony Optimization applied to the Flow Shop Problem

Project on Ant Colony Optimization applied to the Permutation Flow Shop Problem with Weighted Tardiness

Organization

This project is divided into folders:

• instances/ contains the problem instances (i.e. numerical data)
• src/ contains the Python code
• results/ contain the numerical results in CSV files
• plots/ contains the EPS plots of the results used in the report
• report/ contains the LaTeX report

The code is divided into two different types of source files: the algorithm code and the scripts.

The script files are all those that begins with run_ or plot_, and are responsible for running the experiments and exporting the results. They all take two command-line arguments, the first being the input file or folder, the second being the output file or folder. The run_ files take as input the path to the instance file or folder, and outputs a CSV file with numerical results. The plot_ files take as input the path to the CSV result files (or folder thereof) and output an EPS plot.

The algorithm code files are all the others, and implement the various algorithms we use in our experiments.

Requirements

This project is fully written in Python 3. For installing the required libraries, run

pip3 install -r requirements.txt

Running the project

To run the parameter optimization experiment, run in the command line

python3 src/run_parameter_optimization.py instances/ results/

To plot the results of the parameter optimization experiment, run

python3 src/plot_parameter_optimization.py results/<algo name>-results-opti-param.csv plots/<algo name>-params.eps

To run all algorithm 10 times on each instance, run

python3 src/run_performance.py instances/ results/

To plot the results of the runs, run

python3 src/plot_performance.py results/ plots/small-instances-results.eps plots/large-instances-results.eps

To perform the Wilcoxon rank-sum test to compare different algorithms, run

python3 src/run_wilcoxon_text.py results/ results/all-wcn.csv

To run algorithms and plot their convergence, run

python3 src/run_plot_convergence.py instances/<instance name>.txt plots/<instance name>-convergence.eps

In order to keep the code simple, we do not provide a simple script to run the experiment where we the local search is evaluated. Instead, we use the previous scripts by changing the arguments and some of the code. The local search is activated by giving the argument use_local_search = True to the constructor of the RankBasedAS class.