CogniGron Seminar; "Programming with Neurons: Neural Engineering and Neuromorphic Hardware" - Terry Stewart
|When:||Th 05-04-2018 14:00 - 15:00|
|Where:||Energy Academy Building 5159.0062|
Biological systems manage to perform complex calculations using large numbers of low-power spiking components (i.e. neurons). This talk will start by presenting the Neural Engineering Framework; a generic method for combining massively parallel components using weighted connections such that the overall system computes a desired function. This method was originally developed for modelling biological neurons, and we used it to create Spaun, the world's first (and so far only) spiking-neuron brain model capable of performing multiple cognitive tasks. While we have extended this technique to other brain areas and more complex biological neuron models, it has also become clear that this approach is well-suited to developing algorithms that run efficiently on neuromorphic hardware. That is, we can make use of neuromorphic hardware not only to simulate more and more complex brain models, but also to create industry-useful applications for neuromorphic hardware. I will discuss our work on adaptive motor control (enabling systems to learn to deal with physical wear and tear) using Intel's new digital neuromorphic chip Loihi, and our work helping develop Stanford's new analog neuromorphic chip Brainstorm.