Low-energy fuzzy computing by memristors

Nonlinear systems implemented by neural networks represent a paradigm shift in modelling decision-making processes, reaching the limit of the Turing test using high-performance computing. However, compared to the human brain, their energy consumption is a million times worse. Therefore, we study the possibility of replacing the transistor as an elementary logic element with a memristor. The results demonstrate that memristor circuits can implement not only Boolean logic but also Zadeh logic for decision making under uncertainty using filamentary memristors. In cooperation with FY Bizzcom, we implement memristor circuits for fuzzy control and pattern recognition. Such an implementation enables explainability; therefore, the main research directions are memristor implementation of fuzzy logic in decision-making processes and AI explainability.