how to get the triangle subgraph

User 314 | 3/17/2015, 3:24:37 PM

Comments

User 19 | 3/17/2015, 3:28:01 PM

Hi Charley,

Could you please clarify your desired output? Are you using GraphLab Create's <a href="https://dato.com/products/create/docs/generated/graphlab.trianglecounting.create.html#graphlab.trianglecounting.create">triangle counting method</a>?

Thanks, Chris


User 314 | 3/17/2015, 3:45:41 PM

yes I am Graphlab Create's triangle counting method. I want to access the triangle graph with vertices and edge weight, such that I can rank these triangles with some metrics involving edge weights.


User 19 | 3/17/2015, 4:44:37 PM

We do not expose an enumeration of the triangles that were detected. We only return the number of triangles that involve each vertex, which you can obtain with m['triangle_count'].

How do you want to use edge weights to rank the triangles? There may be a way to create a custom triple_apply method to solve your original problem.

Chris


User 314 | 3/17/2015, 5:14:40 PM

thanks. for example, I want to get the average edge weights of each triangle, is there a way to implement using triple_apply? could give me some example code?


User 19 | 3/17/2015, 6:12:37 PM

May I ask why you want to average the edge weights for each triangle? I may be able to suggest a slightly different approach.

It would be a bit tricky to do trianglecount using tripleapply.

How big is your graph?


User 314 | 3/18/2015, 2:14:17 PM

average weights is just one thought, I haven't decided the best way, so it would be good to hear your suggestion. the graph is not big, about one hundred nodes, 2-3 thousand edges.


User 19 | 3/18/2015, 4:06:07 PM

If possible, can you tell me what are you trying to understand about the graph? Knowing a bit about your goals would help me come up with potential solutions.

Thanks! Chris


User 314 | 3/19/2015, 4:57:46 PM

I want to "rank" 3-way collaborations in the graph, and need to consider edge weights.