How does our brain compute better than machines?
Understanding which mechanisms make it so much more capable than the most powerful supercomputer, is a path of potentially profound impact on businesses and economies. The most successful cognitive computing and machine learning systems today do indeed use algorithms loosely inspired by the brain, i.e. artificial neural networks. But these networks do not exploit many of the known operational properties that are unique to biological neurons and synapses which form the so-called Spiking Neural Networks.
Dr. Moraitis's recent work on these algorithms demonstrates how Spiking Neural Networks bring qualitatively unique functionality to cognitive computing, while his current research shows - for the first time - quantitatively better performance in data processing thanks to the targeted application of unique biological properties in tasks that require them. Existing and emerging neuromorphic technologies, like phase-change memristors can emulate spiking algorithms with extremely power-efficient operation, but so far these alorithms required a compromise in performance compared to more conventional algorithms. Dr. Moraitis's work shows how the efficient hardware can be used for unique algorithmic functionality.