Synaptic Scaling 2D: Hebbian Plasticity vs Homeostatic Scaling
2D companion to the 3D synaptic-scaling-homeostatic-plasticity-lab scene: real per-synapse Hebbian (rate-covariance) weight updates on a fast timescale, balanced against a slow multiplicative homeostatic scale factor that pulls the neuron's running-average firing rate back to a target setpoint.
This is the 2D counterpart to the 3D synaptic-scaling-homeostatic-plasticity-lab scene. Eight synapses feed a single neuron, each carrying a fixed correlation with a shared background stimulus. A genuine per-synapse Hebbian rule — potentiation or depression proportional to how well a synapse's own presynaptic drive tracks the neuron's deviation from its target firing rate — runs on a fast timescale, while a slow multiplicative homeostatic scale factor renormalizes every synapse together to pull the running-average firing rate back toward its setpoint, preserving the relative pattern of strong and weak synapses the Hebbian rule just created. Watch the scrolling firing-rate trace against its target line and the live weight bars respond as you change the activity preset or toggle either mechanism off.
2D companion to the 3D synaptic-scaling-homeostatic-plasticity-lab scene: a genuine per-synapse rate-covariance Hebbian update runs on a fast timescale across eight synapses with fixed correlations to a shared stimulus, while a slow multiplicative homeostatic scale factor renormalizes every synapse together to pull the neuron's running-average firing rate back toward its target setpoint.
2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install