A meta-evolutionary genetic algorithm framework where everything is an Individual and capable of evolution. Based on my previous project, Finch.

git clone https://github.com/yourusername/aule.git
cd aule
pip install -e .

Evolve a string to match "hello world":

from aule import Environment, EvaluationLayer, PopulateLayer, SortLayer, CapPopulationLayer
from aule.species import ListGenePool, ListPointCrossoverLayer, ListRandomMutationLayer

TARGET = list("hello world")
VOCAB = list("abcdefghijklmnopqrstuvwxyz ")

genepool = ListGenePool(vocab=VOCAB, length=len(TARGET))
fitness = lambda ind: sum(1 for a, b in zip(ind.genes, TARGET) if a == b)

env = Environment(layers=[
    PopulateLayer(genepool, size=500),
    ListPointCrossoverLayer(points=2, offspring_count=250),
    ListRandomMutationLayer(VOCAB, rate=0.02),
    EvaluationLayer(fitness),
    SortLayer(),
    CapPopulationLayer(500),
])

env.evolve(epochs=100)
print(''.join(env.best_ever.genes))  # hello world

All components—environments, layers, operators—are Individuals that can evolve. This enables co-evolution at every level: mutation strategies compete within mutation layers, layers compete within environments, and environments compete within meta-environments.

Built-in representations for different problem types:

Tensors (aule.species.tensor_evo) - PyTorch tensors for numerical optimization

• TensorIndividual, TensorGenePool
• PointCrossoverLayer, UniformCrossoverLayer, BlendCrossoverLayer
• GaussianMutationLayer, RandomMutationLayer, SwapMutationLayer

Lists (aule.species.list_evo) - Sequences from any vocabulary

• ListIndividual, ListGenePool
• ListPointCrossoverLayer, ListUniformCrossoverLayer
• ListRandomMutationLayer, ListSwapMutationLayer, ListShuffleMutationLayer

Evolution is defined as a pipeline of layers:

• TournamentSelection - Pick best from random subsets
• RouletteSelection - Probability proportional to fitness
• RankSelection - Probability based on rank
• TruncationSelection - Keep top percentage
• ElitistSelection - Preserve top n, tournament for rest

Run the examples:

python examples/hello_world.py      # String evolution
python examples/fitness_comparison.py  # Compare fitness functions