Fracnetics Documentation

Fracnetics is a high-performance C++ library with Python bindings for Genetic Network Programming (GNP) — an evolutionary approach that evolves directed graph structures for decision-making and control tasks.

The Library extends Genetic Network Programming (GNP) with variable-size networks and fractal geometry, designed to model adaptive, recursive, and self-similar network structures inspired by biological evolution and natural growth patterns.

It is designed for researchers and developers exploring:

• Evolutionary network design

• Directed multigraph network architecture

• variable-size networks

• Complex systems with self-similar topology

Contents:

• C++ API Reference
  • Population
  • Network
  • Node
  • Fractal
• Python API Reference
  • Utilities
• 📓 Examples and Tutorials

Installation

` pip install fracnetics `

Architecture Overview

Fracnetics is organized into three core components:

1. Population – Manages selection, mutation, crossover, and generations

2. Network – Directed multigraph with evolutionary operators and fitness functions

3. Node – Basic unit (start, judgment, processing)

Key Features

Population Management

• Create and manage a full population of Network individuals.

• Graph-Based Evolution: Networks evolve as directed graphs with judgment nodes (conditional decisions) and processing nodes (actions), enabling node reuse and cyclic execution paths

Fitness Evaluation

• Built-in support for OpenAI Gymnasium environments

• Traverse a network path via callTraversePath and calculate any own fitness/ target function

Selection & Elitism: Tournament selection with configurable size.

Mutation Operators

• Edge mutation – altering connections between nodes.

• Boundary mutations (multiple variants): - Uniform - Normally distributed - Network-size dependent sigma scaling - Edge-count dependent sigma scaling - Fractal-based mutation

Crossover: Node exchange between individuals.

Innovative Operators

• Add/Delete Nodes – dynamic structural changes in networks.

• Fractal Geometry Integration – hierarchical boundary generation via production rules (L-systems-style subdivision).

Applications

Fracnetics targets domains requiring adaptive, interpretable graph structures:

• Real-time control and robotics

• Classification

• Reinforcement learning problems

• Combinatorial and parameter optimization

• Time series prediction with recurrent connections