Generalizing Wolfram’s Elementary Cellular Automata to Low-Density Networks: A Case Study in Text Classification
O. Johansson and Z. Konkoli
This study focuses on realizing reservoir computing applications using low-density cellular automata networks. The innovation lies in generalizing Wolfram’s elementary nearest-neighbor cellular automata rules to arbitrary network structures, with text classification serving as the test case. The primary objective is to optimize system performance by leveraging the interplay between update rules and network topology. By exploring various network configurations, the study aims to uncover structural characteristics that facilitate efficient information propagation and decision-making within the cellular automata framework. Observations from the study indicate the existence of advantageous combinations of network structures and update rules that enhance reservoir quality. The cellular automata networks considered are categorized into different groups based on their performance, providing a solid foundation for further research.
Keywords: Cellular automata networks, reservoir computing, text classification
DOI: 10.32908/ijuc.v19.220724