Reservoir computing based on cellular automata (ReCA) to solve the 20 QA bAbI Tasks from Facebook

Abstract

Reservoir computing based on cellular automata (ReCA) constructs a bridge between automata computational theory and recurrent neural architectures. Sequence learning is an essential capability for a wide collection of intelligence tasks such as language, continuous vision, and symbolic manipulation in a knowledge base, etc. Therefore, in this study, our model (ReCA) has been developed to solve the 20 QA bAbI tasks from Facebook. These tasks are very hard and require an understanding of the meaning of a text and the ability to reason over relevant facts. ReCA could solve most of the bAbI tasks; 15 out of 20 had 100% accuracy, and 2 tasks were above 90%, whilst 3 tasks were less than 90%, using only the CA evolution states at the last time step as a feature vector to predict the output (LAST method). The CA evolution states at all time steps (ALL method) can also be used, which improves the ReCA accuracy with a large feature space. ReCA results were competitive with the state-of-the-art methods in 2025, such as MT-DNS, and even ReCA outperforms NeuroLogicX, even though these methods are more complicated compared with ReCA.