Handwritten Character Recognition and Image Classification using Neural Cellular Automata

Abstract

Neural Cellular Automata (NCA) are a computational model that incorporates neural networks into the update rules ofcellular automata. In this study, we introduce three different types of NCA modes for handwritten character recognition andimage classification. We evaluate their performance using four different datasets: MNIST, fashionMNIST, CIFAR-10 andETL-8. The experimental results show that the proposed systems successfully recognize handwritten characters and classifyimages. Moreover, it is shown that the NCA models exhibit higher performance than the CNN model when learning fromsmall data.