Transient State Reorganization and Cell Differentiation in the Developmental Dynamics of Growing Neural Cellular Automata
Neural and Evolutionary Computing
Summary
The authors studied how Growing Neural Cellular Automata (GNCA) grow complex shapes from a single cell by examining the full development process of trained models. They found that growth happens through changing temporary stages instead of steady improvement. The hidden parts of the model organize into groups that evolve alongside the visible shape. By analyzing the model’s internal states, the authors discovered distinct cell types appear during growth, which correspond to different regions in the final shape. Overall, the growth process is more about reorganizing these temporary states than simple step-by-step refinement.
Authors
Hiroki Sato, Atsushi Masumori, Takashi Ikegami
Abstract
Growing Neural Cellular Automata (GNCA) develop complex morphologies from a single seed cell through shared local rules, yet the internal dynamics of this process remain poorly understood. To investigate how GNCA grows, the full developmental trajectory of trained GNCA models was traced. The trajectory of cell state development revealed that morphological convergence often proceeds non-monotonically through transient intermediate configurations. In addition, channel-wise analysis showed that the hidden channels self-organize into modular groups in parallel with the visible form. Furthermore, geometric analysis of the cell state space indicated that cell states diversify within a low-dimensional, smooth manifold. To examine cell development in more detail, community detection on an $ε$-neighbour network of cells was conducted. This analysis successfully extracted discrete cell types from this continuous space, and identified transient cell-type communities during early development and stable, finer-grained types corresponding to spatially coherent regions of the mature morphology. The temporal coordination of these phenomena across multiple independent measures indicates that the developmental process of GNCA is a reorganization of transient states rather than incremental refinement.