Final Report Summary - NEUROP (Neuromorphic processors: event-based VLSI models of cortical circuits for brain-inspired computation)
Brains are remarkable computing devices which clearly outperform conventional computers in real-world tasks. Standard machine learning approaches are achieving impressive results with neural networks, but at the cost of extremely high power consumption, using bulky machines and very large data sets. In this project, we designed new analog electronic circuits that use the physics of silicon to directly emulate the biophysics of real neurons and synapses. We used these circuits to build compact and ultra-low power neuromorphic computing systems that reproduce the learning, adaptation, and computational properties of cortical circuits, and developed computational neuroscience methods to program them to carry out procedural tasks. We embedded these neuromorphic processors in autonomous sensory-motor agents and employed them to interact intelligently with the environment in real-time, demonstrating how they can make complex state-dependent computations, and produce behaviors that express cognitive abilities.