Artificial Intelligence (AI) within IoT devices requires an unprecedented amount of data flow and analysis. This is seriously challenging the conventional computation hardware in terms of speed and power consumption, especially if learning capability is required. Therefore, biologically inspired neuromorphic computing systems based on emerging memory devices, which combine storage and processing, are proposed as a viable solution. The FeMTo project works towards a neuromorphic network capable of performing AI tasks while operating at an extremely low power consumption in the order of fJ/bit. The research will include development of new ferroelectric memory devices as well as design of integrated circuits in order to bring neuromorphic computing to a level where proof-of-concept system demonstrates an ultra-low power operation without compromising the computational speed, accuracy, or learning capacity.

2022

2026