Can someone confirm my finding the implementation of floyd_warshall_numpy method of networkx 2.5 is incorrect?
The code to reproduce is:
G = nx.balanced_tree(2, 3)
print(G.nodes())
print(nx.shortest_path(G, 2, 13))
print(nx.floyd_warshall_numpy(G, [2, 8, 13]))
My output is
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14]
[2, 6, 13]
[[ 0. inf inf]
[inf 0. inf]
[inf inf 0.]]
I expected non Inf distances are computed for all [2, 8, 13] node pairs as the shortest path exists among all of them. It seems to me this implementation tries somehow to find path in a subgraph.
nx.floyd_warshall_numpy(G)
works correctly for all nodes. I find the documentation not intuitive here. https://networkx.org/documentation/stable/reference/algorithms/generated/networkx.algorithms.shortest_paths.dense.floyd_warshall_numpy.html#networkx.algorithms.shortest_paths.dense.floyd_warshall_numpy
I've looked at the source code, and what's happening is that the algorithm only considers paths that involve the nodes you've given it.
So since there is no path between 2 and 8 that only includes the nodes 2, 8, and 13, it's returning inf.
I am not sure how best it should be fixed - whether it's better to update the documentation or the method.