Seonglae ChoSpiking neural network
Closely mimic natural neural networks
Artificial neural network that more closely mimics natural neural networks. In addition to neuron and synapse states, SNNs integrate the concept of time into their operational model
Membrane Dynamics
In real neurons, charges accumulate in the cell membrane until they reach a certain threshold (Vₜₕ), at which point they fire an electrical spike. This is mathematically modeled most notably by the Leaky Integrate-and-Fire (LIF) model.
Refractory Period
After firing a spike, neurons cannot fire again for a short period. This period is called the refractory period. This prevents neurons from firing excessively at the same time and allows information to be propagated "temporally sparsely" through the network.
Temporal Coding
SNNs don't represent values as floating-point numbers, but rather by "when the spike occurred". For example, bright pixels trigger quick and frequent spikes, while dark pixels trigger late or almost no spikes. This allows simultaneous use of value and pattern over time.
SNNs
