gsnn.gsnn.models.SignedMessagePassing

class gsnn.gsnn.models.SignedMessagePassing(*args: Any, **kwargs: Any)[source]

Bases: MessagePassing

Aggregate scalar signals over function-function edges using stored signs (edge weights).

__init__(edge_weight, edge_index)[source]
Parameters:

  • edge_weight – Per-edge scalar weights (e.g. ±1), concatenated homogeneous ordering.

  • edge_index – [2, E] homogeneous indices matching edge_weight.

Methods

__init__(edge_weight, edge_index)

param edge_weight:

Per-edge scalar weights (e.g. ±1), concatenated homogeneous ordering.

forward(x)

Propagate x of shape (B, N_fn) over function-only subgraph; returns (B, N_fn).

message(x_i, x_j, edge_weight)

Neighbor contribution weighted by edge_weight.
forward(x)[source]

Propagate x of shape (B, N_fn) over function-only subgraph; returns (B, N_fn).

message(x_i, x_j, edge_weight)[source]

Neighbor contribution weighted by edge_weight.