In the Nengo documentation example for creating a rectified linear neuron, the operator uses updates
:
self.reads = [J]
self.updates = [output]
self.sets = []
self.incs = []
However, in the Nengo reference backend, the neuron models use sets
instead of updates
:
I believe updates
will induce a single time-step delay in the neural response, whereas sets
will take effect at the start of the time-step?