Spike Encoding: Rate, Temporal & Population Coding (2D)

Neuromorphic hardware and spiking neural networks never move a plain floating-point number between units — every value has to be turned into a train of discrete spikes first. This 2D raster view takes one continuous input signal and encodes it live, side by side, with the three encoding schemes used across real neuromorphic systems: rate coding (spike frequency carries the value), temporal/latency coding (spike timing carries the value), and population coding (which neuron in a tuned bank is most active carries the value). Three scrolling spike rasters — one row per neuron — update in real time as spikes are generated from Poisson-style and tuning-curve statistics, while live readouts decode each scheme's estimate of the true signal and track its error and spike-rate cost. Drag to pan the raster and scroll to zoom in on individual spikes, making the classic accuracy-vs-efficiency tradeoff between the schemes directly visible rather than just described.