SNN is more active instead of passive compared to ANN. SNN can work in both modes: Feed & self-thinking (from a point in neuralnet); while ANN is feed-based only. Even more advanced, SNN should be more bio-simulated instead of bio-inspired (maths). Even more bio, multiple types of neurons instead of the common multi-polar neuron, eg. convo, LSTM neurons have multi-polar dendrites but different at core (nucleus).
ChatGPT:
It’s called “spiking” in Spiking Neural Networks (SNNs) because the neurons in these models communicate using discrete events called spikes (or action potentials), rather than continuous values as in traditional artificial neural networks (ANNs).
User: So it's event-based to trigger groups of neurons instead of feeding from input to output?
ChatGPT:
Exactly — you got it 👍
In a standard ANN, data flows layer by layer in a feedforward, clock-like manner:
-
Input → hidden layers → output.
Every neuron updates every step, regardless of whether it’s "important" at that moment.
In a Spiking Neural Network (SNN), it’s event-driven instead:
-
Neurons sit idle most of the time.
They only “fire” (emit a spike) when their membrane potential reaches a threshold due to accumulated input.
That spike is an event that gets transmitted to connected neurons.
Those neurons may (or may not) spike in response, depending on their own internal state.
