NetworkX to_numpy_array parallel edge weight aggregation

Unexpected values in NetworkX to_numpy_array output for parallel edges

0 reputation · 16 Nov 2023, 04:23 UTC

18.4K views0

When converting a NetworkX MultiGraph or MultiDiGraph that contains parallel edges to a NumPy array using to_numpy_array with a weight attribute, the resulting matrix may show values that do not match any single edge weight.

The function’s documentation does not state how it combines the weights of parallel edges, so users cannot determine whether the values represent a sum, maximum, first‑encountered weight, or another aggregation.

What aggregation rule does to_numpy_array currently apply to parallel edges?
Is this rule documented or made configurable through a parameter?
Should downstream numeric analyses assume the current behavior, or should users verify the result by alternative means?

4,520 following

Keep exploring this topic

Graph.nodes() and Graph.edges() changed from lists to views in networkx 2.0

When upgrading to networkx 2.0+, Graph.nodes() and Graph.edges() return view objects instead of lists, breaking code that relies on list‑specific operations.

NetworkX graph serialization and attribute type persistence

Analysis of attribute persistence and type integrity when serializing NetworkX graphs to JSON and GML formats.

TypeError: 'generator' object is not subscriptable when indexing networkx.all_pairs_shortest_path_length result

NetworkX 3.0 changed all_pairs_shortest_path_length to return a generator, causing a TypeError when code tries to index the result as a dictionary.

Explicit weight parameter vs default unweighted mode in NetworkX shortest_path

Discusses the trade‑off between explicitly specifying edge weights and relying on NetworkX’s default unweighted behavior when computing shortest paths.

Unexpected values in NetworkX to_numpy_array output for parallel edges

0 answers

Your perspective belongs here.

0 question comments