Speaker of Workshop 3
Will talk about: Brian for neuromorphic computing
Romain Brette is a research director leading the Computational Neuroscience of Sensory Systems group in the Vision Institute (Paris, France). He is a nominated member of Institut Universitaire de France, and a laureate of the European Research Council. He is the co-editor of the Springer series in computational neuroscience, and a program board member of the OCNS and INCF. He has published about 50 peer-reviewed articles on single neuron models, sensory systems, psychology of perception, acoustics, neural simulation and electrophysiology techniques (including Neuron, eLife, PNAS, J Neurosci, PLoS Comp Biol). He is the co-author of the Brian simulator, the most popular spiking neural network simulator; and of one of the most popular simplified neuron models (the AdEx model). His main interests are neural excitability, spike-based computation, neural computation in ecological environments, auditory and visual perception, neural network simulation.
Neuromorphic computing relies on tools to simulate neuron models on various platforms, including dedicated electronic devices. This diversity of simulation platforms is a challenge. To address this challenge, there is a trend towards the standardization of models. Here I will argue that neuron and plasticity models are likely to evolve quite drastically in the near future, and that standards face the risk of getting rapidly outdated. I will then expose the alternative strategy used in the Brian neural simulator, where models are defined directly by their mathematical equations and code is automatically generated for each specific target.