Release Notes: GraSPy 0.0.2¶
We're happy to announce the release of GraSPy 0.0.2! GraSPy is a Python package for understanding the properties of random graphs that arise from modern datasets, such as social networks and brain networks.
For more information, please visit our website and our tutorials.
Highlights¶
This release is the result of 3 months of work with over 16 pull requests by 5 contributors. Highlights include:
Nonparametric hypothesis testing method for testing two non-vertex matched graphs.
Plotting updates to
pairplot
,gridplot
andheatmaps
.Sampling degree-correlcted stochatic block models (DC-SBM).
import_edgelist
function for importing single or multiple edgelists.Enforcing
Black
formatting for the package.
Improvements¶
Embedding methods are now fully sklearn-compliant. This is tested via
check_estimator
function in sklearn.gridplot
andheatmap
can now plot hierchical labels.New Laplacian computing method ('R-DAD') by adding a constant to the diagonal degree matrix.
Semiparametric testing only checks for largest connected component (LCC) in the intial embeddings.
Various bug fixes.
Various tutorial latex fixes.
Various documentation clarifications.
More consistent documentation.
API Changes¶
check_lcc
argument inAdjacencySpectralEmbed
,LaplacianSpectralEmbed
, andOmnibusEmbed
classes, which checks if input graph(s) are fully connected whencheck_lcc
is True.gridplot
andheatmap
now have ainner_hier_labels
andouter_hier_labels
, which are used for hierarchical labeling of nodes.to_laplacian
function now hasregularizer
arg for whenform
is 'R-DAD'.sbm
function now hasdc
anddc_kws
arguments for sampling SBM with degree-correction.
Deprecations¶
None.