RituPande / hnca

Code for 'Hierarchical Neural Cellular Automata' (Alife 2023)

Home Page:https://direct.mit.edu/isal/proceedings/isal/35/20/116844

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Hierarchical Neural Cellular Automata

This repository contains all necessary code, images, model weights and colab file to reproduce the results in the paper:

Hierachical Neural Cellular Automata
Ritu Pande, Daniele Grattarola

Experiments

There are two experiments performed to show the capabilities of H-NCA, namely, Cell Differentiation and Metamorphosis . Cell Differentiation involves morphing of the patterns generated by the child-NCA to the patterns required by the parent-NCA while Metamorphosis involves descruction of the patterns generated by the child-NCA and the morphology required by the parent-NCA takes over.

Results

Cell Differentiation

cd1 cd2 cd3

Metamomorphosis

m1 m2 m3

Reproduction of Results

To reproduce the results of the paper, open hnca.ipynb from the repository in Google Colab and load the model weights. Then evolve the model using those weights. You can also train the models afresh following the instructions in the notebook.

About

Code for 'Hierarchical Neural Cellular Automata' (Alife 2023)

https://direct.mit.edu/isal/proceedings/isal/35/20/116844


Languages

Language:Jupyter Notebook 98.5%Language:Python 1.5%