add fast_rp algorithm #1867
add fast_rp algorithm #1867
Conversation
Added Rust implementation; Rust test; Python integration; Python test
| }); | ||
| } | ||
|
|
||
| // NOTE(Wyatt): the simple fast_rp test is more of a validation of idempotency than correctness (although the results are expected) |
There was a problem hiding this comment.
Same as this here - wanna create a fastRP 2.0 ticket to track this
There was a problem hiding this comment.
On the point of extra testing, perhaps some window tests + layer tests etc - this is probably a good idea for us to add within the larger test pipeline
There was a problem hiding this comment.
Also yes.
| let weights = Arc::clone(&weights); | ||
| let ss = vv.eval_graph.ss; | ||
| // TODO: rewrite using iters? | ||
| for neighbour in vv.neighbours() { |
There was a problem hiding this comment.
this algorithm is treating the network as undirected, would a directed version make sense?
There was a problem hiding this comment.
The base implementation is only defined for undirected/bidirectional graphs, so this is correct as is. There is, however, a modification I intend to make in the future where different segments of the embedding vector encode different edge types and/or features.
…ordered AlgorithmResult; make tidy NOTE: make tidy applied formatting to: python/tests/graphql/test_nodes_property_filter.py, python/tests/graphql/test_graph_nodes_edges_property_filter.py, and python/python/raphtory/__init__.pyi
NOTE: when running black, some minor formatting changes applied to
python/python/raphtory/graphql/__init__.pyi,python/python/raphtory/__init__.pyi,python/tests/graphql/edit_graph/test_graphql.py,python/tests/test_graphdb/test_graphdb.py, andpython/python/raphtory/algorithms/__init__.pyi(this last one was expected, since it was auto-generated by pyO3)What changes were proposed in this pull request?
Added FastRP to algorithms list. Added: Rust implementation; Rust test; Python integration; Python test
Why are the changes needed?
Always good to expand the algorithms library!
Does this PR introduce any user-facing change? If yes is this documented?
Yes. The Python and Rust functions include docstrings.
How was this patch tested?
The test includes a small sample graph with a seeded result which conforms to the expectations of the embedding algorithm.
Are there any further changes required?
(This goes for all algorithms!) Add expanded validation testing + scale testing.