Neuromorphic circuit design
Published : 19 December 2016
The objective is to develop a Digital Signal Processor (DSP) that is robust and ultra-low-power, targeting mobile and energy-autonomous applications. This Neuro-DSP will be implemented using a spiking neural network. The research activity of this Post-Doc will consist in implementing the base mathematical functions using neurons, deriving from it more complex signal processing functions (eventually using a compiler), determine the size of the neuromorphic circuit and participate to its design.
The candidate will perform simulations using an in-house neuromorphic simulator, called XNET, in order to validate the base concepts and develop the signal processing functions Library. He will also study the obtained accuracy: for example, the noise floor of a Fast Fourier Transform (FFT). Finally, the candidate will size the neuromorphic circuit architecture and participate to its design.