dsmga2 0.2.0

Dependency Structure Matrix Genetic Algorithm II - A fast genetic algorithm with linkage learning

DSMGA2 Python Bindings

Python bindings for DSMGA2 (Dependency Structure Matrix Genetic Algorithm II), a fast genetic algorithm with automatic linkage learning.

Installation

pip install dsmga2

Or build from source:

pip install maturin
maturin develop --release

Quick Start

from dsmga2 import Dsmga2, OneMax

# Create optimizer for 100-bit OneMax problem
optimizer = Dsmga2(problem_size=100, fitness_function=OneMax())
optimizer.population_size = 200
optimizer.max_generations = 1000
optimizer.seed = 42

# Run optimization
result = optimizer.run()
print(f"Best fitness: {result.best_fitness}")
print(f"Generations: {result.generation}")
print(f"Evaluations: {result.num_evaluations}")

Built-in Fitness Functions

• OneMax(): Counts the number of 1s in the chromosome
• MkTrap(k=5): m-k Trap function with deceptive blocks

Custom Fitness Functions

You can implement custom fitness functions in Python:

class CustomFitness:
    def evaluate(self, genes):
        """
        Evaluate fitness of a chromosome.

        Args:
            genes: List of booleans representing the chromosome

        Returns:
            float: Fitness value (higher is better)
        """
        # Your fitness evaluation logic here
        return sum(genes)

    def optimum(self, length):
        """Return the optimal fitness value for this problem size."""
        return float(length)

# Use custom fitness
optimizer = Dsmga2(problem_size=100, fitness_function=CustomFitness())
result = optimizer.run()

API Reference

Dsmga2

Main optimizer class.

Constructor:

Dsmga2(problem_size: int, fitness_function: FitnessFunction)

Attributes:

• population_size: int - Population size (default: problem_size)
• max_generations: int - Maximum generations (default: -1, unlimited)
• seed: int - Random seed (default: 42)

Methods:

• run() -> OptimizationResult - Run optimization until convergence or max generations
• step() -> bool - Run one generation, returns True if not converged
• best_fitness() -> float - Get current best fitness
• generation() -> int - Get current generation number

OptimizationResult

Result object returned by run().

Attributes:

• best_fitness: float - Best fitness found
• generation: int - Number of generations run
• num_evaluations: int - Total fitness evaluations
• mean_fitness: float - Mean population fitness
• converged: bool - Whether algorithm converged

License

MIT OR Apache-2.0