Expose capability to run Louvain with different permutations of vertices to the python layer

[blackboxo]

What is your question?

I want to generate several different results of Louvain, and need to random permute vertices in graph.

In igraph, we can make it like this:
p = np.random.permutation(self.vcount()).tolist()
g = self.permute_vertices(p)

How to realize the same function in multi-GPU cuGraph with dask cudf? Thanks

Code of Conduct

• I agree to follow cuGraph's Code of Conduct
• I have searched the open issues and have found no duplicates for this question

[ChuckHastings]

Is the goal to run Louvain one the same graph with a different ordering of the vertices?

We have a feature in Louvain (to support the ECG algorithm) which will do this for you. Unfortunately, this is not currently exposed to the python layer. Short of that, there's no real good solution to this problem. Vertices are ordered in cugraph based on a hashing/sorting technique that is deterministic. So even if you were to scramble the input and create a new graph you'd end up with essentially the same order.

We could look at exposing this to the python layer.

[blackboxo]

Yes, the goal is to run Louvain one the same graph with a different ordering of the vertices.