Interactive Spike-Timing-Dependent Plasticity (STDP) visualization — explore how precise spike timing shapes synaptic weights, the neural basis of learning and memory
STDP is a biological learning rule discovered by Bi & Poo (1998). When a presynaptic neuron fires shortly before a postsynaptic neuron (Δt > 0), the synapse is strengthened (LTP). When the order reverses (Δt < 0), the synapse weakens (LTD). This temporally asymmetric learning window refines Hebb's postulate into a precise, causally meaningful rule.
The STDP weight change follows: ΔW = A+·exp(−Δt/τ+) for Δt > 0 (LTP), and ΔW = −A−·exp(Δt/τ−) for Δt < 0 (LTD). A+ and A− are learning rate amplitudes, τ+ and τ− are time constants (~20 ms). Weight update: w ← w + ΔW, clamped to [w_min, w_max].
STDP implements causal learning: if presynaptic activity precedes postsynaptic firing, the connection strengthens. If it follows, the connection weakens. This enables networks to learn temporal sequences and form memory engrams.
STDP underlies memory formation in hippocampus, refines circuits during development, powers neuromorphic chips (Intel Loihi), enables autonomous robot learning, and inspires spiking neural network algorithms.
The top panel shows the STDP learning window: red (LTP) for Δt > 0, blue (LTD) for Δt < 0. The vertical line marks your Δt. The middle panel tracks weight over training. The bottom heatmap shows 8 synapses with different Δt values evolving simultaneously.
1) Set Δt = +10 ms and Train — observe LTP. 2) Set Δt = −10 ms — observe LTD. 3) Sweep Δt from −50 to +50. 4) Try 'Asymmetric' preset. 5) Switch to 'Burst' protocol. 6) Try 'Anti-Hebbian' — the curve inverts! 7) Reset and use Step to advance one pair at a time.