- Added density and group connection tests #1032 <#1032>
- Fixed NetworkX 3 compatibility and switched to sparse arrays #1018 <https://github.com/microsoft/graspologic/commit/13d0d466fd1f28c6504b83ae63c16e53c4445997>
- Corrected contributing guidelines #1014 <#1014>
- Corrected deprecation warnings #1019 <#1019>
- Limited NetworkX version due to incompatibility #1016 <#1016>
- Added Python 3.10 and removed Python 3.7 support #1010 <#1010>
- Fixed bug with a matplotlib version incompatibility #996 <#996>
- Fixed graph matching with similarity matrix of unequal dimensions #1002 <#1002>
- Fixed bug with missing typing-extensions dependency #999 <#999>
- Refactored graph matching code and added many new features #960 <#960>
- Added elbow marker to screeplot in plot module #979 <#979>
- Fixed mug2vec behavior for directed graphs #968 <#968>
- Fixed typo in aligning tutorial #974 <#974>
- Added sex labels to mice dataset #967 <#967>
- Made improvements to contributing guidelines #973 <#973>
- Corrected notation in documentation of to_laplacians #969 <#969>
- Fixed isolated nodes handling in node2vec #953 <#953>
- Fixed repeated numba compilation in EdgeSwapper #965 <#965>
- Fixed intersphinx bug #963 <#963>
- Removed default axis labels in networkplot #954 <#954>
- Fixed reproducibility in EdgeSwapper and added to docs #945 <#945>
- Added Degree Preserving Edge Swaps #935 <#935>
- Fixed mypy issue #943 <#943>
- Fixed loops bug in SBM and DCSBM model fitting #930 <#930>
- Added error message in Leiden when given a multigraph was incorrect #926 <#926>
- Fixed typos in ER and SBM models #920 <#920>
- Removed Python 3.6 support
- Officially added Python 3.9 support
- Fixed a type in an error message #904 <#904>
- Added support for arbitrarily indexed node data for networkplot #906 <#906>
- Fixed compatibility issues with gensim, black, mypy, and setup-python #913 <#913>
- Fixed a bug in leiden/hierarchical_leiden #902 <#902>
- Ensured reproducibility in latent_distribution_test #892 <#892>
- Fixed documentation in KMeansCluster #892 <#829>
- Fixed bug in automatic layouts #894 <#894>
- Added Python 3.9 support #889 <#889>
- Added type hinting #543 <#543>
- Modified order of returns in inference module #859 <#859>
- Changes to documentation publish pipeline #874 <#874>
- Added function networkplot for drawing 2D networks #860 <#860>
- Removed dtype requirement in vertex nomination #865 <#865>
- Fixed imports for hyppo >= 0.2.0 #785
- Added
trials
parameter to leiden and set a new requirement ofgraspologic-native>=1.0.0
#790 - Updated lcc check in utils to use
scipy's
connected_components inis_fully_connected
#708 - Added Out of Sample Laplacian Spectral Embedding #722
- Added parallelization of sampling process in Latent Distribution Test #744
- Added Covariate Assisted Spectral Embedding #599
- Added sparse matrix support for some matrix plotting functions #794
- Add
scipy's
connected component finder for sparse support inlargest_connected_component
#795 - Default weight to 1 for all edges in unweighted graph for node2vec #789
- Fixed a bug with 0's in sparse matrix for largest connected component calculation #805
- Usage of the 'un-bearably' awesome beartype library for type checking #819
- Added directed graph support to automated layouts #807
- Fixed bug in mug2vec around
pass_to_ranks
#821 - Pipeline module released, which includes a
networkx
based API for using Adjacency or Laplacian Spectral Embeddings and the Omnibus Spectral Embedding. #814, #817, #823, #824 - Add option for more than one kmeans init to autogmm #662
- Added sparse support for Omnibus embeddings #834
- Added LSE as an embedding for use within the Omnibus embeddings #835
- Clarified behavior of Leiden for graphs with isolates #830
- Updated
utils.is_unweighted
to use more efficient scipy count and filter methods for CSR matrices #836 - Updated default values for node2vec parameters to be more in line with most common production settings #838
- Documentation fixes and updates #729 and #743
- Fixed custom initialization handling for GraphMatch #737
- Fixed incorrect use of optimal transport params in seedless procrustes #745
- Added Similarity Augmented Graph Matching #560
- Added parallelized boostrapping to the latent position test #710
- Fixed outlier handling in quadratic assignment #754
- Fixed GraphMatch random_state behavior when parallelizing in #770
- Modified autolayout to allow you to opt out of the slow no overlap / occlusion removal algorithm
- Fixed bug in leiden where graphs with non-string node IDs were being returned as their
str()
representations instead of their original values. - Fixed a number of bugs in autolayout
- Fixed multiple bugs in MASE, #766 and #768
- Fixed bug in node2vec where non-string node IDs were always being returned as their
str()
representation in the labels array - Fixed bug in autolayout where non-string node IDs broke (see: leiden fix and node2vec fix)
This release represents the first release of the GraSPy
and topologic libraries merger!
In addition to all of the features that GraSPy
had in last-graspy-label
, this release includes a number of feature enhancements and bug fixes - frankly, too many to go into.
Please treat this as a brand new first release, informed heavily by the prior GraSPy
, and extra special thanks to all of our contributors!
release/graspy_releases.rst