GraphEvolutionTool

The Graph Evolution Tool (GET) generates and refines synthetic network data using a genetic algorithm. Users supply a custom fitness function and application-specific parameters, and GET evolves networks meeting those criteria. GET offers two GA representations, which can be stacked to produce higher-fitness networks.

Documentation: https://md12ol.github.io/GraphEvolutionTool/ — what every configuration key means, one page per route, and how to add your own objective. Start at The Pipeline for how a run works, or Configuration for the per-key reference.

Configuring a run

A run is configured either by a TOML file or by typed objects in Python. Both go through the same parser and the same validation, so neither accepts a configuration the other would reject.

• TOML — copy config.example.toml and adjust it. It holds two complete setups, one live and one commented out beneath it; switching between them is commenting one block out and uncommenting the other, not editing scattered keys. What each key means is on the Configuration page.
• Python — build the same thing as objects. Worked examples for all four objectives, both genomes and both evolution strategies are in examples/config_builder.py.

examples/ and config.example.toml live in the repository, not in the wheel. Installing the package gets you the extension module and nothing else, so the links above go to GitHub rather than to a path on your machine.

import get

config = get.Config(
    population_size=200,
    network_size=100,
    crossover_rate=0.9,
    mutation_rate=0.2,
    evolution=get.EvolutionConfig.Generational(num_generations=500),
    scope=get.ScopeConfig.Global(),
    selection=get.SelectionConfig.Tournament(tournament_size=5),
    genome=get.GenomeConfig.EdgeEdit(gene_length=256),
    fitness=get.FitnessConfig.EpiSpread(
        sir=get.SirParams(infection_rate=0.05, num_epidemics=30)
    ),
)
evolver = get.GraphEvolver.from_config(config)

config.to_toml() returns the document that was actually parsed — write it beside your results and the run reproduces verbatim.

Installing

Building the extension module needs maturin: maturin develop for a working copy, or pip install ..