Finch: Evolutionary Algorithm Framework

Finch is a Python framework for implementing evolutionary algorithms. It provides a modular approach to building and experimenting with various evolutionary computation techniques.

Key Features

Modular design with customizable components
Support for different types of genes (float arrays, strings, arrays)
Various selection, crossover, and mutation operators
GPU acceleration support using CuPy
Visualization tools for monitoring evolution progress

Main Components

GenePool: Generates initial populations
- FloatPool, StringPool, ArrayPool, ImagePool
Individual: Represents a single solution in the population
Layer: Defines genetic operators
- Selection layers
- Crossover layers (e.g., N-Point, Uniform)
- Mutation layers (e.g., Gaussian, Uniform, Polynomial, Swap, Inversion, Scramble)
Environment: Manages the evolution process
Competition: Allows comparing multiple evolutionary strategies

Usage

Define your fitness function
Create a GenePool
Set up Layers for selection, crossover, and mutation
Initialize an Environment with your layers and individuals
Run the evolution process

Example


import Finch.layers as layers
from Finch.selectors import *
from Finch.generic import *

def fitness_function(individual):
    return sum(individual.item)

gene_pool = layers.float_arrays.FloatPool(ranges=[[-5, 5]] * 10, length=10, fitness_function=fitness_function)
mutation_selection = RandomSelection(percent_to_select=.1)
crossover_selection = RandomSelection(amount_to_select=2)

# Set up layers
layers = [
    layers.universal_layers.Populate(population=500, gene_pool=gene_pool),
    layers.array_layers.ParentNPoint(selection_function=crossover_selection.select, families=4, children=4),
    layers.float_arrays.GaussianMutation(mutation_rate=0.1, sigma=0.5, selection_function=mutation_selection.select),
    layers.universal_layers.SortByFitness(),
    layers.universal_layers.CapPopulation(1000),
]

env = Environment(layers)
env.compile()
env.evolve(generations=1000)

print(env.best_ever.item)
env.plot()