Gnn

[simondelarue]

Modifications:

(new algorithms, added tests or documentation)

• Adds a Graph Neural Network (GNN) based node classifier.
• Adds a pre-defined Graph Convolutional Network (GCN) layer.
• Adds gradient descent and Adam optimizers.
• Adds related tests and tutorial notebook.

Impacted submodules:

• sknetwork.gnn (new module)

This pull request:

(your PR must match these criteria before review)

• [x] targets the develop branch
• [x] does not decrease code coverage
• [x] passes the tests
• [x] is documented and the documentation build passes
• [x] has PEP8-compliant code and explicit variable naming
• [x] has a tutorial (facultative but recommended)

Any doubts about some technicalities? Do not hesitate to look at the dedicated Wiki.

[codecov[bot]]

Codecov Report

Merging #533 (1ed11c2) into develop (382a21c) will increase coverage by 0.12%. The diff coverage is 99.74%.

@@             Coverage Diff             @@
##           develop     #533      +/-   ##
===========================================
+ Coverage    98.57%   98.70%   +0.12%     
===========================================
  Files          173      188      +15     
  Lines         7320     8109     +789     
===========================================
+ Hits          7216     8004     +788     
- Misses         104      105       +1     

Impacted Files	Coverage Δ
sknetwork/utils/check.py	98.70% <83.33%> (+0.07%)	:arrow_up:
sknetwork/gnn/optimizer.py	97.95% <97.95%> (ø)
sknetwork/__init__.py	100.00% <100.00%> (ø)
sknetwork/clustering/louvain.py	100.00% <100.00%> (ø)
sknetwork/data/tests/test_toy_graphs.py	100.00% <100.00%> (ø)
sknetwork/data/toy_graphs.py	100.00% <100.00%> (ø)
sknetwork/embedding/louvain_embedding.py	100.00% <100.00%> (ø)
sknetwork/gnn/__init__.py	100.00% <100.00%> (ø)
sknetwork/gnn/activation.py	100.00% <100.00%> (ø)
sknetwork/gnn/base.py	100.00% <100.00%> (ø)
... and 14 more

:mega: Codecov can now indicate which changes are the most critical in Pull Requests. Learn more