Line graph and vector centrality functions added #290

goznalo-git · 2023-03-11T18:40:27Z

Hi there,

I really like this library, and I've been using it for a while now. I would like to have these two additions included, and I'm sure others would benefit from them.

The function convert_to_line_graph() takes a Hypergraph and constructs its associated line graph (a pairwise graph whose nodes correspond to the edges of the Hypergraph, and two of those nodes are linked together if their associated edges share at least one node).
The function vector_centrality(), defined in https://doi.org/10.1016/j.chaos.2022.112397, which allows one to extract a vector-like eigenvector centrality of a Hypergraph by using the abovementioned connection to its line graph.

I updated the changelog and the docs, as well as making sure the formatting was fine with isort and black.

best,
Gonzalo

leotrs

Dear Gonzalo, thank you for your contribution! The entire purpose of XGI from its inception was to be a tool for the community by the community, so your involvement is more than welcome. I have reviewed the code and it looks like exactly the kind of addition we are always open to receiving. I have left some comments that can hopefully be solved in a few quick iterations and we'll be sure to merge this after that.

Perhaps another pair of eyes would be useful too @nwlandry @maximelucas.

After all the comments are resolved, it would also be nice to have one or two reference tests for the new features.

Thanks again!

docs/source/api/algorithms/xgi.algorithms.centrality.rst

xgi/convert.py

xgi/algorithms/centrality.py

leotrs · 2023-03-11T20:41:22Z

xgi/algorithms/centrality.py

+
+    D = np.max(list(hyperedge_dims.values()))
+
+    for k in range(2, D + 1):


This loop is first iterating over each possible edge order, then filtering to get all edges of said order, and the iterating over each node of each edge of those edges. So it's actually a four-ly nested loop. I wonder if there is a way to iterate over each edge (and each node in each edge) once?

leotrs · 2023-03-11T20:42:06Z

xgi/algorithms/centrality.py

+                    c_i[node] += LGcent[edge_label_dict[edge]]
+                except IndexError:
+                    raise Exception(
+                        "Nodes must be written with the Pythonic indexing (0,1,2...)"


It's not super clear to me what is mean there by Pythonic indexing. Does this mean that the algorithm does not work for nodes that are labeled with strings? Does it work with nodes that are labeled with non-consecutive integers, or integers not starting at zero?

Co-authored-by: Leo Torres <dleonardotn@gmail.com>

Changed vector_centrality to line_vector_centrality

Co-authored-by: Leo Torres <dleonardotn@gmail.com>

maximelucas · 2023-03-13T10:45:28Z

Thanks a lot @goznalo-git ! This looks like a quality PR.
I'll have a look this week.

xgi/algorithms/centrality.py

maximelucas · 2023-03-15T08:20:36Z

xgi/convert.py

+    edge_label_dict = {
+        tuple(edge): index for index, edge in enumerate(H.edges.members())
+    }


Current code will give new labels to the hyperedges from 0 to M. This suggestion keeps the hyperedges labels, which might be unordered, and strings etc. Check it does not mess up the rest of your code.

Suggested change

edge_label_dict = {

tuple(edge): index for index, edge in enumerate(H.edges.members())

}

edge_label_dict = {tuple(edge): index for index, edge in H._edge}

xgi/convert.py

maximelucas · 2023-03-15T08:42:39Z

xgi/convert.py

+    for edge1, edge2 in combinations(H.edges.members(), 2):
+        if edge1.intersection(edge2):
+            LG.add_edge(edge_label_dict[tuple(edge1)], edge_label_dict[tuple(edge2)])


Your code definitely works. I'm wondering if it might be faster using H.edges.neighbors(). I give it a try below, feel free to discard.

Suggested change

for edge1, edge2 in combinations(H.edges.members(), 2):

if edge1.intersection(edge2):

LG.add_edge(edge_label_dict[tuple(edge1)], edge_label_dict[tuple(edge2)])

edges = []

for edge_id1 in H.edges():

for edge_id2 in H.edges.neighbors(edge_id1):

edges.append({edge_id1, edge_id2})

LG.add_edges_from(edges)

maximelucas · 2023-03-15T08:50:27Z

xgi/algorithms/centrality.py

+    edge_label_dict = {
+        tuple(edge): index for index, edge in enumerate(H.edges.members())
+    }


Do you need labeling from 0 to M? If not see below for suggestion to keep hyperedges labels

maximelucas · 2023-03-15T08:59:25Z

xgi/algorithms/centrality.py

+
+    D = H.edges.size.max()
+
+    for k in range(2, D + 1):


Shouldn't this go up to D since you're iterating over sizes?

maximelucas · 2023-03-15T09:01:04Z

xgi/algorithms/centrality.py

+    for k in range(2, D + 1):
+        c_i = np.zeros(len(H.nodes))
+
+        for edge, _ in list(filter(lambda x: x[1] == k, hyperedge_dims.items())):


Should also work with (check if not slower). Then you don't need to create hyperedge_dims anymore

Suggested change

for edge, _ in list(filter(lambda x: x[1] == k, hyperedge_dims.items())):

for edge in H.edges.filterby("size", k).members():

maximelucas · 2023-03-15T09:44:13Z

Thank you very much @goznalo-git. I added some suggestions to make use of some xgi features.
Thanks a lot!

Two suggestions added, the rest have been replied to. Co-authored-by: Maxime Lucas <maximelucas@users.noreply.github.com>

goznalo-git · 2023-03-17T10:13:35Z

Hi @leotrs @maximelucas

Thanks for your feedback! There were some interesting suggestions and improvements, indeed. Let me know what you think of the actual status.

best!

codecov · 2023-03-17T13:53:43Z

Codecov Report

❗ No coverage uploaded for pull request base (main@d032a3b). Click here to learn what that means.
Patch has no changes to coverable lines.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #290   +/-   ##
=======================================
  Coverage        ?   83.41%           
=======================================
  Files           ?       35           
  Lines           ?     2732           
  Branches        ?        0           
=======================================
  Hits            ?     2279           
  Misses          ?      453           
  Partials        ?        0

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

maximelucas · 2023-03-21T13:46:44Z

Thank you @goznalo-git. I'll approve this for merging now, and raise some issues for our remaining comments so we can solve later together with adding tests.

leotrs · 2023-03-22T09:36:36Z

@goznalo-git, thank you for your contribution. You are more than welcome to continue contributing anything you'd find useful, and to come chat with us whenever on our zulip server.

Line graph and vector centrality functions added

ff53e29

leotrs reviewed Mar 11, 2023

View reviewed changes

goznalo-git and others added 4 commits March 11, 2023 22:19

Update xgi/algorithms/centrality.py

f96edd5

Co-authored-by: Leo Torres <dleonardotn@gmail.com>

Update xgi.algorithms.centrality.rst

37ff20c

Update centrality.py

76b4c38

Changed vector_centrality to line_vector_centrality

Apply suggestions from code review

1ae8113

Co-authored-by: Leo Torres <dleonardotn@gmail.com>

maximelucas reviewed Mar 14, 2023

View reviewed changes

xgi/algorithms/centrality.py Outdated Show resolved Hide resolved

maximelucas reviewed Mar 15, 2023

View reviewed changes

xgi/convert.py Outdated Show resolved Hide resolved

maximelucas reviewed Mar 15, 2023

View reviewed changes

goznalo-git and others added 2 commits March 17, 2023 11:08

Apply suggestions from code review

b167a46

Two suggestions added, the rest have been replied to. Co-authored-by: Maxime Lucas <maximelucas@users.noreply.github.com>

Added fixed suggestion from code review

4b2e328

maximelucas approved these changes Mar 21, 2023

View reviewed changes

maximelucas mentioned this pull request Mar 21, 2023

add tests for #290 #298

Closed

nwlandry merged commit 6a5fb60 into xgi-org:main Mar 21, 2023

maximelucas mentioned this pull request Mar 28, 2023

Centrality update #313

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Line graph and vector centrality functions added #290

Line graph and vector centrality functions added #290

goznalo-git commented Mar 11, 2023

leotrs left a comment

leotrs Mar 11, 2023

leotrs Mar 11, 2023

maximelucas commented Mar 13, 2023

maximelucas Mar 15, 2023

maximelucas Mar 15, 2023

maximelucas Mar 15, 2023

maximelucas Mar 15, 2023

maximelucas Mar 15, 2023

maximelucas commented Mar 15, 2023

goznalo-git commented Mar 17, 2023

codecov bot commented Mar 17, 2023

maximelucas commented Mar 21, 2023

leotrs commented Mar 22, 2023


		D = np.max(list(hyperedge_dims.values()))

		for k in range(2, D + 1):

-    for edge1, edge2 in combinations(H.edges.members(), 2):
-        if edge1.intersection(edge2):
-            LG.add_edge(edge_label_dict[tuple(edge1)], edge_label_dict[tuple(edge2)])
+    edges = []
+    for edge_id1 in H.edges():
+        for edge_id2 in H.edges.neighbors(edge_id1):
+            edges.append({edge_id1, edge_id2})
+    LG.add_edges_from(edges)

	for edge, _ in list(filter(lambda x: x[1] == k, hyperedge_dims.items())):
	for edge in H.edges.filterby("size", k).members():

Line graph and vector centrality functions added #290

Line graph and vector centrality functions added #290

Conversation

goznalo-git commented Mar 11, 2023

leotrs left a comment

Choose a reason for hiding this comment

leotrs Mar 11, 2023

Choose a reason for hiding this comment

leotrs Mar 11, 2023

Choose a reason for hiding this comment

maximelucas commented Mar 13, 2023

maximelucas Mar 15, 2023

Choose a reason for hiding this comment

maximelucas Mar 15, 2023

Choose a reason for hiding this comment

maximelucas Mar 15, 2023

Choose a reason for hiding this comment

maximelucas Mar 15, 2023

Choose a reason for hiding this comment

maximelucas Mar 15, 2023

Choose a reason for hiding this comment

maximelucas commented Mar 15, 2023

goznalo-git commented Mar 17, 2023

codecov bot commented Mar 17, 2023

Codecov Report

maximelucas commented Mar 21, 2023

leotrs commented Mar 22, 2023