Researchers review Neural Cellular Automata and release NCAtorch reference implementation

A paper reviews existing work on Neural Cellular Automata (NCA)—systems that combine learnable artificial neural networks with cellular automata (simple recursive programs) to model complex, self-organizing generative systems.

The authors provide a unified modular framework, notation, and an open-source reference implementation called NCAtorch, building on Stephen Wolfram's 2003 work proposing cellular automata as an alternative to traditional mathematical formalizations like differential equations.

The work demonstrates that NCA can learn complex update rules from data samples, enabling them to model systems that would be difficult to capture with conventional approaches.