Exploring-Flowlenia

Exploring-Flowlenia is a framework for simulating, optimizing, and analyzing Flow-Lenia cellular automata systems. Flow-Lenia is an extension of the Lenia cellular automaton model that incorporates fluid dynamics principles to create more complex and emergent behaviors.

Overview

This project provides tools for:

• Creating and simulating Flow-Lenia systems with different parameters
• Optimizing Flow-Lenia parameters using evolutionary algorithms
• Analyzing the complexity and behaviors of Flow-Lenia patterns
• Automatic discovery of interesting patterns using techniques like IMGEP (Intrinsically Motivated Goal Exploration Processes)

Requirements

This project uses python 3.10.*

Installation

To begin, set up a virtual environment and install the required packages. Then install the flowlenia package in editable mode. You can do this by running the following commands:

pip install -r requirements.txt
pip install -e .

By default, Jax is installed with the cpu version. To install the gpu version, follow the instructions at https://jax.readthedocs.io/en/latest/installation.html

Core Components

The project consists of the following main components:

FlowLenia Simulation

• flowlenia_vanilla.py - The core implementation of the Flow-Lenia model
• flowlenia_params.py - Variants of the Flow-Lenia model with parameter localization

Visualization and Utilities

• vizutils.py - Visualization tools for Flow-Lenia simulations
• simutils.py - Simulation utilities
• utils.py - General utility functions

Optimization

• optimization/evoflow.py - Evolutionary algorithms for optimizing Flow-Lenia parameters
• optimization/tasks.py - Task definitions for evolutionary optimization
• optimization/optexp.py - Experiment configuration for optimization runs

Analysis

• analysis/ - Tools for analyzing Flow-Lenia patterns and behaviors

Auto-discovery

• autodisc/ - Automatic discovery of interesting patterns using IMGEP

Example Usage

See the examples/ directory for examples of how to use the Flow-Lenia framework. Here's a simple example of creating and simulating a Flow-Lenia system:

from flowlenia.flowlenia_vanilla import FlowLenia, Config, State
import jax
import jax.numpy as jnp
import jax.random as jr
import numpy as np
from flowlenia.utils import conn_from_matrix

# Set system configuration
M = np.array([[3, 1, 0], [0, 3, 1], [1, 0, 3]], dtype=int)
k = M.sum()
c0, c1 = conn_from_matrix(M)
cfg = Config(X=128, Y=128, C=3, k=k, c0=c0, c1=c1)

# Initialize system
key = jr.key(1111)
fl = FlowLenia(cfg, key=key)

# Initialize state
s = fl.initialize()
A = s.A.at[44:84, 44:84, :].set(jr.uniform(key, (40, 40, cfg.C)))
s = s.replace(A=A)

# Simulate rollout
s, ss = fl.rollout(s, steps=100)

Experiments

The experiments/ directory contains various experiments showcasing different aspects of Flow-Lenia:

• complexity_analysis/ - Analyzing the complexity of Flow-Lenia patterns
• distribution/ - Experiments with diffusion of matter
• ea_walls/ - Evolutionary activity with wall constraints
• large_map/ - Simulations with large grid sizes
• metamorph/ - Experiments with metamorphosing patterns
• mixing/ - Parameter mixing experiments