More than Moore with Neuromorphic Computing Architectures

Neuromorphic computing systems represent a departure from the von Neumann architecture. They implement aspects of form and function observed in biological neural circuits. Potential advantages include energy efficiency, fault tolerance, compactness and, most importantly, the ability to learn by interaction with the environment. Applications are in two areas : Improving the understanding of biological systems and cognitive computing to analyze causal relations and complex data and to make predictions. In the colloquium I will review current approaches in neuromorphic computing and discuss current and future use cases. Reading suggestions for the enthusiastic public: Learning : https://arxiv.org/abs/1604.05080 Mixing : https://arxiv.org/abs/1709.08166 Noise : https://arxiv.org/abs/1710.04931 Dendrites : https://arxiv.org/abs/1703.07286 Some experiments : https://arxiv.org/abs/1703.06043 Sampling : https://arxiv.org/abs/1311.3211