Spike-Timing-Dependent Plasticity

Neuromorphic chips such as Intel's Loihi learn on-chip, without backpropagation, using local synaptic rules that only look at the relative timing of spikes. This simulator renders one postsynaptic neuron wired to twelve presynaptic inputs split into two groups — six that fire together as a synchronized volley, and six that fire independently in the background — and runs a real trace-based spike-timing-dependent plasticity (STDP) rule on every synapse in real time. Watch the correlated group's synapses (orange) strengthen while the uncorrelated background synapses (blue) weaken, purely from local timing statistics, and tune the firing rates and learning rate to see how the effect strengthens, weakens, or reverses.