Leniabreeder

Repository for "Toward Artificial Open-Ended Evolution within Lenia using Quality-Diversity" (ALIFE 2024).

Installation

git clone https://github.com/maxencefaldor/Leniabreeder.git && cd Leniabreeder

Using virtual environment

At the root of the repository, execute:

python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Using container

At the root of the repository, execute:

apptainer build \
	--fakeroot \
	--force \
	--warn-unused-build-args \
	apptainer/container.sif apptainer/container.def

Run Experiments

Using virtual environment

At the root of the repository, execute:

source venv/bin/activate

Using container

At the root of the repository, execute:

apptainer shell \
	--bind $(pwd):/workspace/src/ \
	--cleanenv \
	--containall \
	--home /tmp/ \
	--no-home \
	--nv \
	--pwd /workspace/src/ \
	--workdir apptainer/ \
	apptainer/container.sif

Commands

To run an experiment with the default configuration, execute the following command:

python main.py seed=$RANDOM qd=<algo>

with <algo> replaced with either me or aurora.

All hyperparameters are available in the configs/ directory and can be overridden via the command line. For example, to run the MAP-Elites experiments as described in the paper, use:

python main.py seed=$RANDOM qd=me qd.n_generations=4_000 qd.repertoire_size=32_000 qd.fitness=pos_linear_velocity_avg qd.descriptor=[color] qd.descriptor_min=[0.,0.,0.] qd.descriptor_max=[1.,1.,1.]

Analyze Experiments

When you run an experiment, a directory is created in output/. To analyze the results, you can either run a script from the analysis/ directory or use the notebooks from the notebooks/directory. Don't forget to change run_dir to the path of your experiment.

For MAP-Elites, you can use analysis/visualize_me.py or notebooks/visualize_me.ipynb. For AURORA, you can use analysis/visualize_aurora.py or notebooks/visualize_aurora.ipynb.

BibTeX

@article{faldor2024leniabreeder,
	author    = {Faldor, Maxence and Cully, Antoine},
	title     = {Toward Artificial Open-Ended Evolution within Lenia using Quality-Diversity},
	journal   = {Artificial Life},
	year      = {2024},
}