Propagation

APPNP spreading activation, STDP edge updates, and LIF pressure decay.

compute_propagation

compute_propagation(graph: DiGraph, source_id: str, grade: Grade, plasticity: Plasticity) -> dict[str, float]

Run APPNP propagation from a fired neuron, return pressure deltas.

Parameters:

Name	Type	Description	Default
graph	DiGraph	The circuit's NetworkX DiGraph.	required
source_id	str	The neuron that was just reviewed.	required
grade	Grade	The review grade (affects propagation strength).	required
plasticity	Plasticity	Plasticity parameters.	required

Returns:

Type	Description
dict[str, float]	Dict of {neuron_id: pressure_delta} for all affected neurons.
dict[str, float]	The source neuron is NOT included (it gets reset separately).

compute_stdp

compute_stdp(graph: DiGraph, fired_id: str, grade: Grade, fired_at: datetime, plasticity: Plasticity) -> list[tuple[str, str, float, bool]]

Compute STDP weight updates for edges adjacent to the fired neuron.

For each edge connecting fired_id to a neighbor: - If neighbor fired recently (within tau_stdp), compute timing-dependent weight change using exponential STDP window. - Pre→post (LTP): Δw = +a_plus * exp(-|Δt| / tau_stdp) - Post→pre (LTD): Δw = -a_minus * exp(-|Δt| / tau_stdp)

MISS grade does not trigger co-fire or LTP (only LTD if applicable).

Returns:

Type	Description
list[tuple[str, str, float, bool]]	List of (pre, post, new_weight, is_co_fire) for each updated edge.

decay_all_pressure

decay_all_pressure(graph: DiGraph, now: datetime, plasticity: Plasticity) -> None

Apply LIF leak to all neuron pressures.

Pressure decays exponentially: u(t) = u * exp(-dt / tau_m)