Differentiable Programming of Reaction-Diffusion Patterns

TL;DR

This paper introduces a differentiable optimization approach to learn parameters of reaction-diffusion systems for texture synthesis, enabling more systematic and effective pattern generation compared to traditional trial-and-error methods.

Contribution

It presents a novel differentiable programming framework for reaction-diffusion systems, integrating neural cellular automata and task-specific loss functions for improved pattern synthesis.

Findings

Generated RD patterns exhibit robust, life-like behaviors.

The method outperforms traditional trial-and-error approaches.

Enables systematic parameter learning for complex pattern formation.

Abstract

Reaction-Diffusion (RD) systems provide a computational framework that governs many pattern formation processes in nature. Current RD system design practices boil down to trial-and-error parameter search. We propose a differentiable optimization method for learning the RD system parameters to perform example-based texture synthesis on a 2D plane. We do this by representing the RD system as a variant of Neural Cellular Automata and using task-specific differentiable loss functions. RD systems generated by our method exhibit robust, non-trivial 'life-like' behavior.