Hi Baisyet,
I think that Jason has/had a narrow view of what constitutes neuromorphic, being analog neurons.
It is true that analog neurons more closely resemble biological neurons by adding actual voltages and firing when a threshold voltage value is reached. A leaky-integrate-and-fire (LIF) neuron can be represented by a capacitor in parallel with a resistor. As the capacitor is charged by input current spikes from other neurons, it begins to discharge (leak) through the resistor, so the rate of incoming spikes is important. If the rate is too low, the capacitor voltage will drop faster than it is built up by the spikes and the neuron will not reach the threshold.
Akida is designed to produce a digital imitation of the function of biological neurons. Akida's neurons "count" digital bits and fire when the count reaches a numerical threshold. The digital output of Akida is equivalent to the result of passing the output of an analog neuron through an analog-to-digital converter, which is a necessary step with an analog neuron if its output is to be useful in a CPU/GPU. Akida uses a time window to replicate the leak by only counting the most recent block of N spikes.
Akida is suitable for edge devices because of its speed of response (low latency) and its low power requirements which make it ideal for battery operated devices.
