Graph Evolution Tool · GET

Design networks by evolving them

GET searches for a graph that performs well on a goal you define. It repeatedly builds candidate networks, measures them, and breeds the better candidates. Use it from Python, a TOML config file, or, when working from the source tree, Rust.

Choose Your Starting Point

I want to use GET

Run a small experiment

Build the Python module, evolve a network, and inspect the result.

Getting Started → I want to understand it

Learn the ideas first

Graphs, genomes, objectives, selection, variation, and what the loop is doing.

Concepts & Vocabulary → I want to contribute

Find the right extension point

Set up the repository, run its checks, and follow the map to the code you need.

Contributor Guide →

What GET Does

You supply an experiment: how many nodes the graph has, how candidates are represented, and the number that says whether one graph is better than another. GET handles the evolutionary search.

1. Genomea compact recipe
2. Graphthe network it builds
3. Fitnessyour objective's score
4. Evolveselect, change, and repeat

The graph is the thing you care about. The genome is the form GET can mutate and breed.

You do not need prior knowledge of evolutionary algorithms or graph theory to begin. The concepts page introduces the terms; the deeper pipeline guide explains the implementation.

What Goes In and What Comes Back

You provide	GET provides
Graph size and optional starting graph	The best graph as weighted edge triples
A built-in or custom objective	The winning score and genome representation
Population, selection, mutation, and stopping settings	A convergence history for every logged iteration
A master seed and number of replicates	Run metadata and saveable results; preserve the code, inputs, and environment too

An Actual Result

An empty 40-node starting graph beside the evolved result, which has 89 edges. — An edge-edit run starts with no edges and evolves a network selected for epidemic spread. This is measured output, not a mock-up.

Where GET Fits

GET is useful when

You can score a candidate network

Your score might measure epidemic behavior, similarity to reference networks, robustness, cost, or a domain-specific property written in Python or Rust.

GET does not decide

What biological meaning is valid

For a gene, protein, metabolic, or contact network, you define what nodes, edges, and a good score mean. GET optimizes that definition; it does not infer biological truth.

Check the data model

Fixed nodes and undirected edges

GET uses integer-indexed, attribute-free graphs with dense storage. Read Data & Inputs before adapting a biological network.

Next Step

If you want a working result, continue to Getting Started or modify the complete files in the Example Bundle. If you already know the package and need a particular interface, use Choose a Route.