Add eigenvector centrality function #621

mtreinish · 2022-06-04T15:59:47Z

This commit adds a new function eigenvector_centrality() to compute the
eigenvector centrality of a graph. It uses the same power function
approach that the NetworkX function eigenvector_centrality() [1]
function uses. This is for two reasons, the first is that a more
traditional eigenvector linear algebra/BLAS function would either
require us to link against BLAS at build time (which is a big change
in the build system and a large requirement) or to call out to numpy
via python both of which seemed less than ideal. The second reason was
to make handling holes in node indices bit easier. Using this approach
also enabled us to put the implementation in retworkx-core so it can be
reused with any petgraph graph.

Part of #441

This commit adds a new function eigenvector_centrality() to compute the eigenvector centrality of a graph. It uses the same power function approach that the NetworkX function eigenvector_centrality() [1] function uses. This is for two reasons, the first is that a more traditional eigenvector linear algebra/BLAS function would either require us to link against BLAS at build time (which is a big change in the build system and a large requirement) or to call out to numpy via python both of which seemed less than ideal. The second reason was to make handling holes in node indices bit easier. Using this approach also enabled us to put the implementation in retworkx-core so it can be reused with any petgraph graph. Part of Qiskit#441

coveralls · 2022-06-04T16:23:30Z

Pull Request Test Coverage Report for Build 2477501415

109 of 120 (90.83%) changed or added relevant lines in 3 files are covered.
2 unchanged lines in 1 file lost coverage.
Overall coverage increased (+0.1%) to 97.254%

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
retworkx-core/src/centrality.rs	37	38	97.37%
src/centrality.rs	68	78	87.18%

Files with Coverage Reduction	New Missed Lines	%
src/shortest_path/all_pairs_bellman_ford.rs	2	98.88%

Totals
Change from base Build 2428334538:	0.1%
Covered Lines:	12394
Relevant Lines:	12744

💛 - Coveralls

retworkx-core/src/centrality.rs

src/centrality.rs

georgios-ts · 2022-06-07T09:46:06Z

src/centrality.rs

+    let cost_fn = CostFn::try_from((weight_fn, default_weight))?;
+    let ev_centrality = centrality::eigenvector_centrality(
+        &graph.graph,
+        |e| cost_fn.call(py, e.weight()),


It might be beneficial to precompute and store all edge weights like we do in all_pairs_dijkstra_path_lengths so we avoid calling Python overhead each iteration.

Let me benchmark this to see if it makes a difference (we only did the precomputation in the all pairs functions because it operates in parallel and we didn't want to be gil bound in parallel threads)

I benchmarked this and it's basically the same but marginally faster in a couple of cases:

not really a noticeable difference. ~~But I think I'll make the change anyway, mostly because it does open up potential parallelization. I think I can make the inner loop execute in parallel and will test that in bit.~~ (I played with this a bit and I don't think it's actually feasible without locking which would defeat the purpose of doing it in parallel)

Oh, actually ignore that graph I forgot to set a weight callback function in the benchmark so that's improvement is just the time for using default weight. Which still seems worthwhile to me

We might want to benchmark against a more "difficult" graph since a path graph will converge after 12 iterations (I think) so the performance difference will be really minor.

I pushed this up in: 3204e09 I'm still waiting on the data collection for the run with a python callback. It's been running the sweep for ~60min and the precomputed weights sweep took ~15min. So there is no doubt this is faster. I'll generate the graph when it eventually finishes just for posterity.

This is the graph with a callback set (I just used float).

src/centrality.rs

retworkx-core/src/centrality.rs

Co-authored-by: georgios-ts <[email protected]>

This commit fixes an issue where we were overcounting parallel edges in the graph. There was a bug in the logic that was calculating the combined weight of parallel edges incorrectly. Co-authored-by: georgios-ts <[email protected]>

This commit changes the callback from the retworkx functions for determining edge weight to precompute them to avoid requiring a python context in the callback. This make a minor improvement in runtime since we only callback to python once at the very beginning instead of in the middle of the loop.

georgios-ts

Thanks for all the updates. Some minor comments but it's good to go.

retworkx-core/src/centrality.rs

src/centrality.rs

georgios-ts

Oh, we'll need to update the docs before we merge.

Co-authored-by: georgios-ts <[email protected]>

georgios-ts requested changes Jun 7, 2022

View reviewed changes

mtreinish and others added 8 commits June 8, 2022 16:04

Fix issues with docstrings

54a9ed2

Co-authored-by: georgios-ts <[email protected]>

Correct default value of max_iter

81e4e2f

Co-authored-by: georgios-ts <[email protected]>

Apply suggestions from code review

e910939

Co-authored-by: georgios-ts <[email protected]>

Fix compiler warning

7798859

Switch to a Vec<f64> instead of HashMap to track centrality

09595a1

Update retworkx-core tests too

97fcc61

Fix handling of parallel edges

e03856b

This commit fixes an issue where we were overcounting parallel edges in the graph. There was a bug in the logic that was calculating the combined weight of parallel edges incorrectly. Co-authored-by: georgios-ts <[email protected]>

mtreinish requested a review from georgios-ts June 10, 2022 14:10

georgios-ts approved these changes Jun 10, 2022

View reviewed changes

retworkx-core/src/centrality.rs Outdated Show resolved Hide resolved

src/centrality.rs Outdated Show resolved Hide resolved

src/centrality.rs Outdated Show resolved Hide resolved

georgios-ts requested changes Jun 10, 2022

View reviewed changes

mtreinish and others added 3 commits June 10, 2022 15:18

Apply suggestions from code review

2f279bd

Co-authored-by: georgios-ts <[email protected]>

Update src/centrality.rs

be031be

Co-authored-by: georgios-ts <[email protected]>

Add docs to toc

9a6ea1a

georgios-ts approved these changes Jun 11, 2022

View reviewed changes

georgios-ts added the automerge Queue a approved PR for merging label Jun 11, 2022

mergify bot merged commit a9c5648 into Qiskit:main Jun 11, 2022

mtreinish deleted the eigenvector-centrality branch June 11, 2022 14:14

IvanIsCoding mentioned this pull request May 26, 2023

Add PageRank #788

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add eigenvector centrality function #621

Add eigenvector centrality function #621

mtreinish commented Jun 4, 2022

coveralls commented Jun 4, 2022 •

edited

Loading

georgios-ts Jun 7, 2022

mtreinish Jun 9, 2022 •

edited

Loading

mtreinish Jun 9, 2022 •

edited

Loading

mtreinish Jun 9, 2022

georgios-ts Jun 9, 2022

mtreinish Jun 9, 2022 •

edited

Loading

mtreinish Jun 9, 2022 •

edited

Loading

georgios-ts left a comment

georgios-ts left a comment

Add eigenvector centrality function #621

Add eigenvector centrality function #621

Conversation

mtreinish commented Jun 4, 2022

coveralls commented Jun 4, 2022 • edited Loading

Pull Request Test Coverage Report for Build 2477501415

💛 - Coveralls

georgios-ts Jun 7, 2022

Choose a reason for hiding this comment

mtreinish Jun 9, 2022 • edited Loading

Choose a reason for hiding this comment

mtreinish Jun 9, 2022 • edited Loading

Choose a reason for hiding this comment

mtreinish Jun 9, 2022

Choose a reason for hiding this comment

georgios-ts Jun 9, 2022

Choose a reason for hiding this comment

mtreinish Jun 9, 2022 • edited Loading

Choose a reason for hiding this comment

mtreinish Jun 9, 2022 • edited Loading

Choose a reason for hiding this comment

georgios-ts left a comment

Choose a reason for hiding this comment

georgios-ts left a comment

Choose a reason for hiding this comment

coveralls commented Jun 4, 2022 •

edited

Loading

mtreinish Jun 9, 2022 •

edited

Loading

mtreinish Jun 9, 2022 •

edited

Loading

mtreinish Jun 9, 2022 •

edited

Loading

mtreinish Jun 9, 2022 •

edited

Loading