















Spiky backpropagation
Neurocomputing
The backpropagation algorithm implemented on spiking neuromorphic hardware
Spiking neural networks, designed to mimic biological brains, could be faster and more efficient than their digital counterparts but traditional training methods have been challenging to implement on neuromorphic hardware that supports them. Here, we demonstrate ‘on-chip’ learning by successfully executing the backpropagation algorithm, which computes weights between nodes, on Intel’s Loihi neuromorphic processor.