Apply event-driven synaptic plasticity

Plasticity rules update synaptic weights in response to spikes. brainevent provides event-driven plasticity operators that touch only the weights connected to neurons that actually fired — the same sparsity principle as the matrix products. This guide shows when to reach for each operator; the full, runnable derivation is in Tutorial 5: Synaptic Plasticity Modeling - Foundation of Learning and Memory.

Pre- vs post-synaptic updates

Plastic updates are driven either by pre-synaptic spikes (the source neuron fired) or post-synaptic spikes (the target neuron fired). brainevent ships both directions for each storage format:

Storage	Pre-synaptic driven	Post-synaptic driven
CSR (sparse)	update_csr_on_binary_pre()	update_csr_on_binary_post()
Dense	update_dense_on_binary_pre()	update_dense_on_binary_post()

Choosing CSR vs dense

• Use the CSR operators when connectivity is sparse and fixed (the common case for large networks) — only the stored synapses are visited.

• Use the dense operators for small, fully-connected layers where a dense weight matrix is already in play.

In both cases the spike trigger is a BinaryArray, so only active events contribute to the update — the cost scales with the number of spikes, not the number of synapses.

See also

The exact operator signatures (weights, spike vector, learning-rate arguments, and the associated *_p primitives) are listed under Plasticity operations in Matrix Operations.