3D Synaptic Plasticity: STDP Volume

Stimulation patterns

Single spike-pair test

Synapse

Δt = t_post − t_pre +10 ms

STDP parameters

Potentiation A₊ 0.12

Depression A₋ 0.12

τ₊ (potentiation) 20 ms

τ₋ (depression) 20 ms

Homeostatic decay 0.010

Weak synapse Mid strength Strong (potentiated) Neuron just fired

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.

Δw vs Δt (STDP curve)

Curve = current A₊/A₋/τ rule · dots = real applied updates

Weight vs time

tracking N0→N1

Spikes

Avg weight

–

Last Δt

–

Last Δw

–

How it works

Ten leaky integrate-and-fire neurons occupy two five-neuron rings stacked in 3D — a front layer and a back layer — wired into a loop: each ring is internally connected neighbour-to-neighbour, every front neuron feeds its counterpart directly behind it, and every back neuron feeds the next neuron around the front ring. A spike can therefore circle the whole volume. Every synapse is rendered as a beam whose thickness and color track its weight, and every connected pair that fires within a short window updates that weight by spike-timing-dependent plasticity (STDP), exactly as in the 2D version of this lab — only here the causal loop is a real path through 3D space that you can orbit around.

Δt = t_post − t_pre
Δw =  A₊·exp(−Δt/τ₊)   if Δt > 0  (pre before post → potentiate)
Δw = −A₋·exp( Δt/τ₋)   if Δt < 0  (post before pre → depress)

Train forward loop fires the whole path in its natural direction, potentiating every synapse it travels through in turn.
Train reverse loop fires the same neurons in reverse order, depressing those same synapses.
Spontaneous activity gives every neuron random background input; once the loop is potentiated, a spike anywhere on it is far more likely to complete a full circuit, visibly lighting the loop up beam by beam.
Drag to orbit, scroll to zoom — the synapse beams' radius and brightness always reflect the live weight, so a fully potentiated loop reads as a bright, thick ring through the volume.

Real-world relevance: this is the same LTP/LTD mechanism believed to underlie memory formation in the hippocampus and cortex, shown here as a genuine 3D circuit so the geometry of a "trained pathway" is visible directly, not just implied by a flat diagram.