minkowski

Module containing EdgeDefinitions based on the Minkowski Metric.

graphnet.models.graphs.edges.minkowski.compute_minkowski_distance_mat(x, y, c, space_coords, time_coord)[source]

Compute all pairwise Minkowski distances.

Parameters:

  • x (Tensor) – First tensor of shape (n, d).

  • y (Tensor) – Second tensor of shape (m, d).

  • c (float) – Speed of light, in scaled units.

  • space_coords (Optional[List[int]], default: None) – Indices of space coordinates.

  • time_coord (Optional[int], default: 3) – Index of time coordinate.

Return type:

Tensor

Returns: Matrix of shape (n, m) of all pairwise Minkowski distances.

class graphnet.models.graphs.edges.minkowski.MinkowskiKNNEdges(*args, **kwargs)[source]

Bases: EdgeDefinition

Builds edges between most light-like separated.

Initialize MinkowskiKNNEdges.

Parameters:

  • nb_nearest_neighbours (int) – Number of neighbours to connect to.

  • c (float) – Speed of light, in scaled units.

  • time_like_weight (float, default: 1.0) – Preference to time-like over space-like edges. Scales time_like distances by this value, before finding nearest neighbours.

  • space_coords (Optional[List[int]], default: None) – Coordinates of x, y, z.

  • time_coord (Optional[int], default: 3) – Coordinate of time.

  • args (Any)

  • kwargs (Any)

Return type:

object