EventProjection

EventProjection#

class brainpy.state.network.EventProjection(*args, **kwargs)#

Delayed, weighted delta-event projection from one population segment.

Each step it reads the pre population’s captured spike via pre_spike() (a callable returning the full pre-population spike/counts vector), applies an InputDelay (full-delay convention, as in AlignPostProj), restricts to this projection’s pre segment, maps it to the post segment (dense weighted matmul, or element-wise for one_to_one), and registers the result as a delta input on post.

Parameters:

  • pre_spike (Callable[[], jax.Array]) – Returns the full pre-population spike (or generator counts) vector, shape (n_pre_pop,).

  • n_pre_pop (int) – Size of the full pre population (the dimension pre_spike returns).

  • pre_local_idx (jax.Array) – Local indices into the pre population selected by this projection.

  • post (Dynamics) – Post-synaptic population; receives add_delta_input.

  • post_local_idx (jax.Array) – Local indices into the post population targeted by this projection.

  • rule (ConnRule) – Connection rule. one_to_one triggers the element-wise path.

  • weight (ArrayLike or Quantity) – Synaptic weight in pA (signed: positive excitatory, negative inhibitory).

  • delay (ArrayLike or Quantity or None) – Axonal delay; None for instantaneous delivery.

  • as_current (bool, default False) – Deposit mode. When False (default) the contribution is registered as a delta input (post.add_delta_input), matching how current-based neurons ingest spikes. When True it is registered as a current input (post.add_current_input) instead — the route the astrocyte slow-inward current (SIC) takes, since SIC is a pA current entering dV/dt rather than a delta/conductance. Requires comm='dense' (a graded current cannot ride the binarising sparse event matmul).

realized_edges()[source]#

Enumerate this projection’s realized edges (NEST GetConnections view).

Returns:

Population-local source / target plus live weight / delay in the canonical edge order, with guarded write-back hooks. See event_proj_edges().

Return type:

ProjEdges