This repository contains the source code for all the experiments in the following paper:
-
The paper is accepted at the Genetic and Evolutionary Computation Conference (GECCO 2023).
-
It is also accepted as a finalist for the Human-Competitive Awards (Humies 2023).
- All the codes for sweeping the parameter space are in the
parameter sweep
folder. RunplotAll.py
for Figure 3 of the paper.
All the codes for the results in section 2.4 of the paper are in the exp1 - single nand
folder.
-
Run
evolveAfpo.py
to redo the experiments. -
Run
afpoPlots.py
andplotInOut.py
for Figures 4, 5 and 6 of the paper. -
Code for the frequency sweep experiment from Figure 7 is in the
frequency sweep
folder.
- Run
plotHeatmap.py
from theexp2 - heatmaps
folder to recreate the heatmap in Figure 8.
Folder exp3 - polycomputation
contains all the codes from section 4 of the paper.
-
Run
evolveAfpo.py
to redo the experiments. -
Run
afpoPlots.py
andplotInOut.py
for Figures 9 and 10 of the paper. -
Folder
exp4 - heatmaps
contains codes to recreate Figure 11 showing the distribution of polycomputation.
- Run
robustness.py
to redo the experiment andplotRobustness.py
to make Figure 12.
evolveAfpo.py
launches an evolutionary algorithm (Age-Fitness Pareto Optimization) with two or three objectives.constants.py
contains the algorithm's parameters such as population size and number of generations.switch_float.py
is the main simulator file. It simulators the 2D granular material taking as input the stiffness matrix and some gate parameters such as the positions of the ports, frequency of vibrations, and phase offset.
@article{parsa2023universal, title={Universal Mechanical Polycomputation in Granular Matter}, author={Parsa, Atoosa and Witthaus, Sven and Pashine, Nidhi and O'Hern, Corey S and Kramer-Bottiglio, Rebecca and Bongard, Josh}, journal={arXiv preprint arXiv:2305.17872}, year={2023} }