scikit-network

• Add Python 3.13
• Add wheel for aarch64
• Fix bug on color display

• Upgrade Numpy

• Add Python 3.12
• Drop Python 3.8

• Add Leiden clustering algorithm
• Add k-center clustering algorithm
• Add functions to detect and break cycles
• Add damping factor in diffusion
• Fix F1 scores
• Remove hierarchical Louvain embedding
• Get clustering coefficient for directed graphs

• Add Leiden clustering algorithm

• Add k-center clustering algorithm

• Add functions to detect and break cycles

• Add damping factor in diffusion

• Fix F1 scores

• Remove hierarchical Louvain embedding

• Get clustering coefficient for directed graphs

• Add Python 3.11
• Add set_param / get_param to algorithms, suggested by Franz Kiraly (#557)
• Compute shortest paths by matrix-vector multiplications
• Make tools on topology (cliques, code-decomposition, etc.) as functions
• Rename parameter membership -> probs for soft classification / clustering
• Add softmax to classification by diffusion

• Add overview
• Add predict_proba method to classification and clustering

• Change API for clustering (predict / transform)
• Change API for classification (seeds -> labels)
• Change API for ranking (seeds -> weights)
• Change API for GNN (same as classifiers)
• Remove first order methods for link prediction
• Add nearest neighbor methods for link prediction
• Add KNN classifier without embedding
• Add TF-IDF
• Solve security issues by upgrade of wheels and ipython

• Drop Python 3.7
• Update Numpy requirement

• Allow Python 3.7 (now working on Colab forcing numpy in 1.22)
• Fix data format on KMeans

• Fix sampling for GraphSage
• Fix leakage on GNNs
• Update tutorial on GNNs with node inference

• Update Graph neural networks (e.g., add GraphSAGE)
• Clean data home folder (set one folder per dataset collection, NetSet, Konect, ...)
• Improve classification by diffusion, setting -1 to unreached nodes
• Fix bug on modularity
• Clean up source distribution
• Safe file extraction
• Fix bug on dendrogram cut
• Add a function to aggregate a graph per label

• Fix documentation

• Drop Python 3.7
• Add graph neural networks
• Clarify predict / transform methods
• Improve classification by diffusion
• Add classification scores
• Add acyclic test for undirected graphs
• Make the algorithms work for all sparse matrix formats

• Add module on regression
• Add connected components for bipartite graphs
• Update functions for loading graphs
• Add radius and eccentricity metrics

• Apple Silicon and Python 3.10 wheel
• Updated parsers
• Updated block models
• Updated spectral embedding

• Merge Bi* algorithms (e.g., BiLouvain -> Louvain) by Thomas Bonald (#490)
• Transition from Travis to Github actions by Quentin Lutz (#488)
• Added sdist build for conda recipes
• Added name position for graph visualization
• Removed randomized algorithms

• Updated NumPy and SciPy requirements

• New push-based implementation of PageRank by Wenzhuo Zhao (#475)
• Fixed cut_balanced in hierarchy
• Dropped Python 3.6, wheels for Python 3.9 (switched to manylinux2014)

• Added hierarchical Louvain embedding by Quentin Lutz (#468)
• Doc fixes and updates
• Requirements update

• Added random projection embedding by Thomas Bonald (#461)
• Added PCA-based embedding by Thomas Bonald (#461)
• Added 64-bit support for Louvain by Flávio Juvenal (#450)
• Added verbosity options for dataset loaders
• Fixed Louvain embedding
• Various doc and tutorial updates

• Added betweenness algorithm by Tiphaine Viard (#444)

• Added Louvain-based embedding
• Fix documentation with new dataset website URLs

Tentative doc fix. The issue appears to come from ReadTheDocs

• Fix visualization features
• Fix documentation

Last release before Nathan's defense ;)

• Added link prediction module
• Added pie-node visualization of memberships
• Added Weisfeiler-Lehman graph coloring
• Added Force Atlas 2 graph layout
• Added triangle listing algorithm for directed and undirected graph
• Added k-core decomposition algorithm
• Added k-clique listing algorithm
• Added color map option in visualization module
• Updated NetSet URL

• Added Katz centrality
• Refactor connectivity module into paths and topology
• Refactor Diffusion into Dirichlet
• Added parsers for adjacency list TSV and GraphML
• Added shortest paths and distances

• Add clustering by label propagation
• Add models
• Add function to build graph from edge list
• Change a parameter in SVG visualization functions
• Minor corrections

Package version for JMLR paper.

• Refactor basics module into connectivity
• Cython version for label propagation
• Minor corrections

• Clarified requirements
• Minor corrections

• Added OpenMP support for all platforms

• Updated ranking module : new pagerank solver, new HITS params, post-processing
• Polynomes in linear algebra
• Added meta.name attribute for Bunch
• Minor corrections

• Added spring layout in embedding

• Added label propagation in classification

• Added save / load functions in data

• Added display edges parameter in svg graph exports

• Corrected typos in documentation

• Minor bug

• Added heat kernel based node classifier
• Updated loaders for WikiLinks
• Fixed file-related issues for Windows

• Added VerboseMixin for verbosity features
• Added Loaders for WikiLinks & Konect databases

• sknetwork: new API for bipartite graphs
• new module: Soft node classification
• new module: Node classification
• new module: data (merge toy graphs + loader)
• clustering: Spectral Clustering
• ranking: new algorithms
• utils: K-neighbors
• hierarchy: Spectral Ward
• data: loader (Vital Wikipedia)

• Minor bug

• Change parser output
• Fix bugs in ranking algorithms (zero-degree nodes)
• Add notebooks
• Import algorithms from scipy (shortest path, connected components, bfs/dfs)
• Change SVD embedding (now in decreasing order of singular values)

• Minor bug

• Changed Louvain, BiLouvain, Paris and PageRank APIs
• Changed PageRank method
• Documentation overhaul
• Improved Jupyter tutorials

• Added Algorithm class for nicer repr of some classes
• Added Jupyter notebooks as tutorials in the docs
• Minor fixes

• Updated PageRank
• Added tests for Numba versioning

• Minor bug

• Unified Louvain.

We added a version of Louvain for directed graphs and a new one for bipartite graphs. A new bipartite toy graph is also available.

Updated clustering module and new documentation.

• First real release on PyPI.

• First release on PyPI.