Pareto Set Plotter

Using generational snapshots, this script creates visual representations of the progression of Pareto optimal sets created by many-objective ant colony optimizers.

Dependencies

• Python3.5+
• PIL (Python Imaging Library)
• ffmpeg: Used to create MP4s, and possibly GIFs, of each visualization. The MP4 versions usually have substantially smaller filesizes, although are in the YUV colour space.
• [Recommended] imagemagick: ImageMagick's convert tool is used to create highly-optimized GIFs. The resulting GIFs are usually substantially smaller than those created using ffmpeg.

Usage

Run this script from within a folder containing generational .pos files, possibly created from my iMOACO_R Implementation; note that you must run iMOACOR.py with the option --snapshots to create the generational .pos snapshots, which will be written to the snapshots directory.

Run with the following command:

python3 <path to plotter.py> [OPTIONS]

Options:

• -d N | --duration=N: The duration in seconds of the output media (default: 5.0)
• -s N | --stepping=N: Only process a filename if its generation number is divisible by N
• -h: Attempt to include hypervolume data for each generation, if available
• --help: Display the help page

For generational hypervolumes to be displayed, they must first be created in the same folder by the use of my Hypervolume Manager. Make sure to use the same or compatible stepping values for each script.

By default, the colour of each solution is determined by its relative rank; to support this, each Pareto set solution must include an additional entry containing the solution's rank.

Examples

PyTorch – ASF: DTLZ4 in 3 dimensions
PyTorch – VADS: DTLZ4 in 3 dimensions
PyTorch – ASF: DTLZ7 in 7 dimensions