Release Notes¶
Release 0.3.1¶
Major Features and Improvements¶
The latest changes made to the
ripser.py
submodule have been pulled (#530, see also #532). This includes in particular the performance improvements to the C++ backend submitted by Julian Burella Pérez via scikit-tda/ripser.py#106. The developer installation now includes a new dependency in robinhood hashmap. These changes do not affect functionality.The example notebook classifying_shapes.ipynb has been modified and improved (#523).
The tutorial previously called
time_series_classification.ipynb
has been split into an introductory tutorial on the Takens embedding ideas (topology_time_series.ipynb) and an example notebook on gravitational wave detection (gravitational_waves_detection.ipynb) which presents a time series classification task (#529).The documentation for
PairwiseDistance
has been improved (#525).
Bug Fixes¶
Timeout deadlines for some of the
hypothesis
tests have been increased to make them less flaky (#531).
Backwards-Incompatible Changes¶
Due to poor support for
brew
in the macOS 10.14 virtual machines by Azure, the CI for macOS systems is now run on 10.15 virtual machines and 10.14 is no longer supported by the wheels (#527)
Thanks to our Contributors¶
This release contains contributions from many people:
Julian Burella Pérez, Umberto Lupo, Lewis Tunstall, Wojciech Reise, and Rayna Andreeva.
We are also grateful to all who filed issues or helped resolve them, asked and answered questions, and were part of inspiring discussions.
Release 0.3.0¶
Major Features and Improvements¶
This is a major release which adds substantial new functionality and introduces several improvements.
Persistent homology of directed flag complexes via pyflagser
¶
The
pyflagser
package (source, docs) is now an official dependency ofgiotto-tda
.The
FlagserPersistence
transformer has been added togtda.homology
(#339). It wrapspyflagser.flagser_weighted
to allow for computations of persistence diagrams from directed or undirected weighted graphs. A new notebook demonstrates its use.
Edge collapsing and performance improvements for persistent homology¶
GUDHI C++ components have been updated to the state of GUDHI v3.3.0, yielding performance improvements in
SparseRipsPersistence
,EuclideanCechPersistence
andCubicalPersistence
(#468).Bindings for GUDHI’s edge collapser have been created and can now be used as an optional preprocessing step via the optional keyword argument
collapse_edges
inVietorisRipsPersistence
and ingtda.externals.ripser
(#469 and #483). Whencollapse_edges=True
, and the input data and/or number of required homology dimensions is sufficiently large, the resulting runtimes for Vietoris–Rips persistent homology are state of the art.The performance of the Ripser bindings has otherwise been improved by avoiding unnecessary data copies, better managing the memory, and using more efficient matrix routines (#501 and #507).
New transformers and functionality in gtda.homology
¶
The
WeakAlphaPersistence
transformer has been added togtda.homology
(#464). LikeVietorisRipsPersistence
,SparseRipsPersistence
andEuclideanCechPersistence
, it computes persistent homology from point clouds, but its runtime can scale much better with size in low dimensions.VietorisRipsPersistence
now accepts sparse input whenmetric="precomputed"
(#424).CubicalPersistence
now accepts lists of 2D arrays (#503).A
reduced_homology
parameter has been added to all persistent homology transformers. WhenTrue
, one infinite bar in the H0 barcode is removed for the user automatically. Previously, it was not possible to keep these bars in the simplicial homology transformers. The default is alwaysTrue
, which implies a breaking change in the case ofCubicalPersistence
(#467).
Persistence diagrams¶
A
ComplexPolynomial
feature extraction transformer has been added (#479).A
NumberOfPoints
feature extraction transformer has been added (#496).An option to normalize the entropy in
PersistenceEntropy
according to a heuristic has been added, and anan_fill_value
parameter allows to replace any NaN produced by the entropy calculation with a fixed constant (#450).The computations in
HeatKernel
,PersistenceImage
and in the pairwise distances and amplitudes related to them has been changed to yield the continuum limit whenn_bins
tends to infinity;sigma
is now measured in the same units as the filtration parameter and defaults to 0.1 (#454).
New curves
subpackage¶
A new curves
subpackage has been added to preprocess, and extract features from, collections of multi-channel curves such as returned by BettiCurve
, PersistenceLandscape
and Silhouette
(#480). It contains:
A
StandardFeatures
transformer that can extract features channel-wise in a generic way.A
Derivative
transformer that computes channel-wise derivatives of any order by discrete differences (#492).
New metaestimators
subpackage¶
A new metaestimator
subpackage has been added with a CollectionTransformer
meta-estimator which converts any transformer instance into a fit-transformer acting on collections (#495).
Images¶
Time series¶
TakensEmbedding
is now a new transformer acting on collections of time series (#460).The former
TakensEmbedding
acting on a single time series has been renamed toSingleTakensEmbedding
transformer, and the internal logic employed in itsfit
for computing optimal hyperparameters is now available via atakens_embedding_optimal_parameters
convenience function (#460).The
_slice_windows
method ofSlidingWindow
has been made public and renamed intoslice_windows
(#460).
Graphs¶
GraphGeodesicDistance
has been improved as follows (#422):The new parameters
directed
,unweighted
andmethod
have been added.The rules on the role of zero entries, infinity entries, and non-stored values have been made clearer.
Masked arrays are now supported.
A
mode
parameter has been added toKNeighborsGraph
; as inscikit-learn
, it can be set to either"distance"
or"connectivity"
(#478).List input is now accepted by all transformers in
gtda.graphs
, and outputs are consistently either lists or 3D arrays (#478).Sparse matrices returned by
KNeighborsGraph
andTransitionGraph
now have int dtype (0-1 adjacency matrices), and are not necessarily symmetric (#478).
Mapper¶
Pullback cover set labels and partial cluster labels have been added to Mapper node hovertexts (#445).
The functionality of
Nerve
andmake_mapper_pipeline
has been greatly extended (#447 and #456):Node and edge metadata are now accessible in output
igraph.Graph
objects by means of theVertexSeq
andEdgeSeq
attributesvs
andes
(respectively). Graph-level dictionaries are no longer used.Available node metadata can be accessed by
graph.vs[attr_name]
where forattr_name
is one of"pullback_set_label"
,"partial_cluster_label"
, or"node_elements"
.Sizes of intersections are automatically stored as edge weights, accessible by
graph.es["weight"]
.A
"store_intersections"
keyword argument has been added toNerve
andmake_mapper_pipeline
to allow to store the indices defining node intersections as edge attributes, accessible viagraph.es["edge_elements"]
.A
contract_nodes
optional parameter has been added to bothNerve
andmake_mapper_pipeline
; nodes which are subsets of other nodes are thrown away from the graph when this parameter is set toTrue
.A
graph_
attribute is stored duringNerve.fit
.
Two of the
Nerve
parameters (min_intersection
and the newcontract_nodes
) are now available in the widgets generated byplot_interactive_mapper_graph
, and the layout of these widgets has been improved (#456).ParallelClustering
andNerve
have been exposed in the documentation and ingtda.mapper
’s__init__
(#447).
Plotting¶
A
plot_params
kwarg is available in plotting functions and methods throughout to allow user customisability of output figures. The user must pass a dictionary with keys"layout"
and/or"trace"
(or"traces"
in some cases) (#441).Several plots produced by
plot
class methods now have default titles (#453).Infinite deaths are now plotted by
plot_diagrams
(#461).Possible multiplicities of persistence pairs in persistence diagram plots are now indicated in the hovertext (#454).
plot_heatmap
now accepts boolean array input (#444).
New tutorials and examples¶
The following new tutorials have been added:
Topology of time series, which explains the theory of the Takens time-delay embedding and its use with persistent homology, demonstrates the new
API
of several components ingtda.time_series
, and shows how to construct time series classification pipelines ingiotto-tda
by partially reproducing arXiv:1910:08245.Topology in time series forecasting, which explains how to set up time series forecasting pipelines in
giotto-tda
viaTransformerResamplerMixin``s and the ``giotto-tda
Pipeline
class.Topological feature extraction from graphs, which explains what the features extracted from directed or undirected graphs by
VietorisRipsPersistence
,SparseRipsPersistence
andFlagserPersistence
are.Classifying handwritten digits, which presents a fully-fledged machine learning pipeline in which cubical persistent homology is applied to the classification of handwritten images from he MNIST dataset, partially reproducing arXiv:1910.08345.
Utils¶
Bug Fixes¶
A bug has been fixed which could lead to features with negative lifetime in persistent homology transformers when
infinity_values
was set too low (#339).By relying on
scipy
’sshortest_path
instead ofscikit-learn
’sgraph_shortest_path
, some errors in computingGraphGeodesicDistance
(e.g. when som edges are zero) have been fixed (#422).A bug in the handling of COO matrices by the
ripser
interface has been fixed (#465).A bug which led to the incorrect handling of the
homology_dimensions
parameter inFiltering
has been fixed (#439).An issue with the use of
joblib.Parallel
, which led to errors when attempting to runHeatKernel
,PersistenceImage
, and the corresponding amplitudes and distances on large datasets, has been fixed (#428 and #481).A bug leading to plots of persistence diagrams not showing points with negative births or deaths has been fixed, as has a bug with the computation of the range to be shown in the plot (#437).
A bug in the handling of persistence pairs with negative death values by
Filtering
has been fixed (#436).A bug in the handling of
homology_dimension_ix
(now renamed tohomology_dimension_idx
) in theplot
methods ofHeatKernel
andPersistenceImage
has been fixed (#452).A bug in the labelling of axes in
HeatKernel
andPersistenceImage
plots has ben fixed (#453 and #454).PersistenceLandscape
plots now show all homology dimensions, instead of just the first (#454).A bug in the computation of amplitudes and pairwise distances based on persistence images has been fixed (#454).
Silhouette
now does not create NaNs when a subdiagram is trivial (#454).CubicalPersistence
now does not create pairs with negative persistence wheninfinity_values
is set too low (#467).Warnings are no longer thrown by
KNeighborsGraph
whenmetric="precomputed"
(#506).A bug in
Labeller.resample
affecting cases in whichn_steps_future >= size - 1
, has been fixed (#460).A bug in
validate_params
, affecting the case of tuples of allowed types, has been fixed (#502).
Backwards-Incompatible Changes¶
The minimum required versions from most of the dependencies have been bumped. The updated dependencies are
numpy >= 1.19.1
,scipy >= 1.5.0
,joblib >= 0.16.0
,scikit-learn >= 0.23.1
,python-igraph >= 0.8.2
,plotly >= 4.8.2
, andpyflagser >= 0.4.1
(#457).GraphGeodesicDistance
now returns either lists or 3D dense ndarrays for compatibility with the homology transformers - By relying onscipy
’sshortest_path
instead ofscikit-learn
’sgraph_shortest_path
, some errors in computingGraphGeodesicDistance
(e.g. when som edges are zero) have been fixed (#422).The output of
PairwiseDistance
has been transposed to matchscikit-learn
convention(n_samples_transform, n_samples_fit)
(#420).plot
class methods now return figures instead of showing them (#441).Mapper node and edge attributes are no longer stored as graph-level dictionaries,
"node_id"
is no longer an available node attribute, and the attributesnodes_
andedges_
previously stored byNerve.fit
have been removed in favour of agraph_
attribute (#447).The
homology_dimension_ix
parameter available in some transformers ingtda.diagrams
has been renamed tohomology_dimensions_idx
(#452).The base of the logarithm used by
PersistenceEntropy
is now 2 instead of e, and NaN values are replaced with -1 instead of 0 by default (#450 and #474).The outputs of
PersistenceImage
,HeatKernel
and of the pairwise distances and amplitudes based on them is now different due to the improvements described above.Weights are no longer stored in the
effective_metric_params_
attribute ofPairwiseDistance
,Amplitude
andScaler
objects when the metric is persistence-image–based; only the weight function is (#454).The
homology_dimensions_
attributes of several transformers have been converted from lists to tuples. When possible, homology dimensions stored as parts of attributes are now presented as ints (#454).gaussian_filter
(used to make heat– and persistence-image–based representations/pairwise distances/amplitudes) is now called withmode="constant"
instead of"reflect"
(#454).The default value of
order
inAmplitude
has been changed from2.
toNone
, giving vector instead of scalar features (#454).The meaning of the default
None
forweight_function
inPersistenceImage
(and inAmplitude
andPairwiseDistance
whenmetric="persistence_image"
) has been changed from the identity function to the function returning a vector of ones (#454).Due to the updates in the GUDHI components, some of the bindings and Python interfaces to the GUDHI C++ components in
gtda.externals
have changed (#468).Labeller.transform
now returns a 1D array instead of a column array (#475).PersistenceLandscape
now returns 3D arrays instead of 4D ones, for compatibility with the newcurves
subpackage (#480).By default,
CubicalPersistence
now removes one infinite bar in H0 (#467, and see above).The former
width
parameter inSlidingWindow
andLabeller
has been replaced with a more intuitivesize
parameter. The relation between the two is:size = width + 1
(#460).clusterer
is now a required parameter inParallelClustering
(#508).The
max_fraction
parameter inFirstSimpleGap
andFirstHistogramGap
now indicates the floor ofmax_fraction * n_samples
; its default value has been changed fromNone
to1
(#412).
Thanks to our Contributors¶
This release contains contributions from many people:
Umberto Lupo, Guillaume Tauzin, Julian Burella Pérez, Wojciech Reise, Lewis Tunstall, Nick Sale, and Anibal Medina-Mardones.
We are also grateful to all who filed issues or helped resolve them, asked and answered questions, and were part of inspiring discussions.
Release 0.2.2¶
Major Features and Improvements¶
The documentation for
gtda.mapper.utils.decorators.method_to_transform
has been improved.A table of contents has been added to the theory glossary.
The theory glossary has been restructured by including a section titled “Analysis”. Entries for l^p norms, L^p norms and heat vectorization have been added.
The project’s Azure CI for Windows versions has been sped-up by ensuring that the locally installed boost version is detected.
Several python bindings to external code from GUDHI, ripser.py and Hera have been made public: specifically,
from gtda.externals import *
now gives power users access to:bottleneck_distance
,wasserstein_distance
,ripser
,SparseRipsComplex
,CechComplex
,CubicalComplex
,PeriodicCubicalComplex
,SimplexTree
,WitnessComplex
,StrongWitnessComplex
.
However, these functionalities are still undocumented.
The
gtda.mapper.visualisation
andgtda.mapper.utils._visualisation
modules have been thoroughly refactored to improve code clarity, add functionality, change behaviour and fix bugs. Specifically, in figures generated by bothplot_static_mapper_graph
andplot_interactive_mapper_graph
:The colorbar no longer shows values rescaled to the interval [0, 1]. Instead, it always shows the true range of node summary statistics.
The values of the node summary statistics are now displayed in the hovertext boxes. A a new keyword argument
n_sig_figs
controls their rounding (3 is the default).plotly_kwargs
has been renamed toplotly_params
(see “Backwards-Incompatible Changes” below).The dependency on
matplotlib
’srgb2hex
andget_cmap
functions has been removed. As no other component ingiotto-tda
requiredmatplotlib
, the dependency on this library has been removed completely.A
node_scale
keyword argument has been added which can be used to controls the size of nodes (see “Backwards-Incompatible Changes” below).The overall look of Mapper graphs has been improved by increasing the opacity of node colors so that edges do not hide them, and by reducing the thickness of marker lines.
Furthermore, a
clone_pipeline
keyword argument has been added toplot_interactive_mapper_graph
, which when set toFalse
allows the user to mutate the input pipeline via the interactive widget.The docstrings of
plot_static_mapper_graph
,plot_interactive_mapper_graph
andmake_mapper_pipeline
have been improved.
Bug Fixes¶
A CI bug introduced by an update to the XCode compiler installed on the Azure Mac machines has been fixed.
A bug afflicting Mapper colors, which was due to an incorrect rescaling to [0, 1], has been fixed.
Backwards-Incompatible Changes¶
The keyword parameter
plotly_kwargs
inplot_static_mapper_graph
andplot_interactive_mapper_graph
has been renamed toplotly_params
and has now slightly different specifications. A new logic controls how the information contained inplotly_params
is used to update plotly figures.The function
get_node_sizeref
ingtda.mapper.utils.visualization
has been hidden by renaming it to_get_node_sizeref
. Its main intended use is subsumed by the newnode_scale
parameter ofplot_static_mapper_graph
andplot_interactive_mapper_graph
.
Thanks to our Contributors¶
This release contains contributions from many people:
Umberto Lupo, Julian Burella Pérez, Anibal Medina-Mardones, Wojciech Reise and Guillaume Tauzin.
We are also grateful to all who filed issues or helped resolve them, asked and answered questions, and were part of inspiring discussions.
Release 0.2.1¶
Major Features and Improvements¶
The theory glossary has been improved to include the notions of vectorization, kernel and amplitude for persistence diagrams.
The
ripser
function ingtda.externals.python.ripser_interface
no longer uses scikit-learn’spairwise_distances
whenmetric
is'precomputed'
, thus allowing square arrays with negative entries or infinities to be passed.check_point_clouds
ingtda.utils.validation
now checks for square array input when the input should be a collection of distance-type matrices. Warnings guide the user to correctly setting thedistance_matrices
parameter.force_all_finite=False
no longer means accepting NaN input (only infinite input is accepted).VietorisRipsPersistence
ingtda.homology.simplicial
no longer masks out infinite entries in the input to be fed toripser
.The docstrings for
check_point_clouds
andVietorisRipsPersistence
have been improved to reflect these changes and the extra level of generality forripser
.
Bug Fixes¶
The variable used to indicate the location of Boost headers has been renamed from
Boost_INCLUDE_DIR
toBoost_INCLUDE_DIRS
to address developer installation issues in some Linux systems.
Backwards-Incompatible Changes¶
The keyword parameter
distance_matrix
incheck_point_clouds
has been renamed todistance_matrices
.
Thanks to our Contributors¶
This release contains contributions from many people:
Umberto Lupo, Anibal Medina-Mardones, Julian Burella Pérez, Guillaume Tauzin, and Wojciech Reise.
We are also grateful to all who filed issues or helped resolve them, asked and answered questions, and were part of inspiring discussions.
Release 0.2.0¶
Major Features and Improvements¶
This is a major release which substantially broadens the scope of giotto-tda
and introduces several improvements.
The library’s documentation has been greatly improved and is now hosted via GitHub pages.
It includes rendered jupyter notebooks from the repository’s examples
folder, as well as an improved theory glossary,
more detailed installation instructions, improved guidelines for contributing, and an FAQ.
Plotting functions and plotting API¶
This version introduces built-in plotting capabilities to giotto-tda
. These come in the form of:
a new
plotting
subpackage populated with plotting functions for common data structures;a new
PlotterMixin
and a class-level plotting API based on newly introducedplot
,transform_plot
andfit_transform_plot
methods which are now available in several ofgiotto-tda
’s transformers.
Changes and additions to gtda.homology
¶
The internal structure of this subpackage has been changed. ConsistentRescaling
has been moved to a new point_clouds
subpackage (see below), and gtda.homology
no longer contains a point_clouds
submodule. Instead, it contains two
submodules, simplicial
and cubical
. simplicial
contains the VietorisRipsPersistence
class as well as the
following new classes:
SparseRipsPersistence
,EuclideanCechPersistence
.
The cubical
submodule contains CubicalPersistence
, a new class for computing persistent homology of filtered cubical
complexes such as those coming from 2D or 3D greyscale images.
New images
subpackage¶
The new gtda.images
subpackage contains classes which, together with gtda.homology.CubicalPersistence
, extend
the capabilities of giotto-tda
to computer vision, by handling input representing binary or greyscale 2D/3D images
represented as arrays.
The classes in gtda.images.filtrations
are responsible for converting binary image input into greyscale images in a
variety of ways. The greyscale output can then be fed to gtda.homology.CubicalPersistence
to extract topological
signatures in the form of persistence diagrams. These classes are:
HeightFiltration
,RadialFiltration
,DilationFiltration
,ErosionFiltration
,SignedDistanceFiltration
.
The classes in gtda.images.preprocessing
perform a variety of preprocessing steps on either binary or greyscale image
input, as well as conversion to point cloud format. They are:
Binarizer
,Inverter
,Padder
,ImageToPointCloud
.
New point_clouds
subpackage¶
ConsistentRescaling
is no longer placed in gtda.homology
. Instead, it is now in a point_clouds
subpackage
containing classes which process or modify the geometry of point cloud data. gtda.point_clouds
also contains the new
class ConsecutiveRescaling
, written with time series applications in mind.
List of point cloud input¶
All classes in the homology
subpackage (VietorisRipsPersistence
, SparseRipsPersistence
, and EuclideanCechPersistence
)
can now take as inputs to the fit
and transform
methods lists of 2D arrays instead of simply 3D arrays. In this
way, collections of point clouds with varying numbers of points can be processed.
Changes and additions to gtda.diagrams
¶
The diagrams
subpackage contains the following new classes:
PersistenceImage
Silhouette
Additionally, the subpackage has been reorganised as follows:
The
features
submodule now only contains the scalar feature generation classesAmplitude
(moved there fromdistance
) andPersistenceEntropy
.Classes which produce vector representations from persistence diagrams have been moved to the new
representations
submodule.
Changes and additions to gtda.utils
¶
validate_params
has been thoroughly refactored, documented and exposed for the benefit of developers.check_diagrams
has been modified, documented and exposed for the benefit of developers.The new
check_point_clouds
performs validation of inputs consisting of collections of point clouds of distance matrices. It accepts both lists of 2D ndarrays and 3D ndarrays, and is used in thefit
andtransform
methods of classes ingtda.homology.simplicial
to allow for list input (see above).
External modules and HPC improvements¶
A substantial effort has been put in improving the quality of the high-performance components contained in gtda.externals
.
The end result is a cleaner packaging as well as faster execution of C++ functions due to improved bindings. In particular:
Two binaries are now shipped for
ripser
, one of them being optimised for calculations with mod 2 coefficients.Recent improvements by the authors of the
hera
C++ library have been integrated ingiotto-tda
.Compiler optimisations for Windows-based systems have been added.
The integration of
pybind11
has been improved and several issues arising withCMake
andboost
during developer installations have been addressed.
Bug Fixes¶
Fixed a bug with
TakensEmbedding
’s algorithm for search of optimal parameters.Inconsistencies in between the meaning of “bottleneck amplitude” in the theory and in the code have been ironed out. The code has been modified to agree with the theory glossary. The outputs of the
gtda.diagrams
classesAmplitude
,Scaler
andFiltering
is affected.Fixed bugs affecting color normalization in Mapper graph plots.
Backwards-Incompatible Changes¶
Python 3.5 is no longer supported.
Mac OS X versions below 10.14 are no longer supported by the wheels shipped via PyPI.
ConsistentRescaling
is no longer found ingtda.homology
and is now part ofgtda.point_clouds
.The outputs of the
gtda.diagrams
classesAmplitude
,Scaler
andFiltering
have changed due to sqrt(2) factors (see Bug Fixes).The
meta_transformers
module has been removed.The
plotting
module has been removed from theexamples
folder of the repository.
Thanks to our Contributors¶
This release contains contributions from many people:
Umberto Lupo, Guillaume Tauzin, Wojciech Reise, Julian Burella Pérez, Roman Yurchak, Lewis Tunstall, Anibal Medina-Mardones, and Adélie Garin.
We are also grateful to all who filed issues or helped resolve them, asked and answered questions, and were part of inspiring discussions.
Release 0.1.4¶
Library name change¶
The library and GitHub repository have been renamed to giotto-tda
! While the
new name is meant to better convey the library’s focus on Topology-powered
machine learning and Data Analysis, the commitment to seamless integration with
scikit-learn
will remain just as strong and a defining feature of the project.
Concurrently, the main module has been renamed from giotto
to gtda
in this
version. giotto-learn
will remain on PyPI as a legacy package (stuck at v0.1.3)
until we have ensured that users and developers have fully migrated. The new PyPI
package giotto-tda
will start at v0.1.4 for project continuity.
Short summary: install via
python -m pip install -U giotto-tda
and import gtda
in your scripts or notebooks!
Change of license¶
The license changes from Apache 2.0 to GNU AGPLv3 from this release on.
Major Features and Improvements¶
Added a
mapper
submodule implementing the Mapper algorithm of Singh, Mémoli and Carlsson. The main tools are the functionsmake_mapper_pipeline
,plot_static_mapper_graph
andplot_interactive_mapper_graph
. The first creates an object of classMapperPipeline
which can be fit-transformed to data to create a Mapper graph in the form of anigraph.Graph
object (see below). TheMapperPipeline
class itself is a simple subclass of scikit-learn’sPipeline
which is adapted to the precise structure of the Mapper algorithm, so that aMapperPipeline
object can be used as part of even larger scikit-learn pipelines, inside a meta-estimator, in a grid search, etc. One also has access to other important features of scikit-learn’sPipeline
, such as memory caching to avoid unnecessary recomputation of early steps when parameters involved in later steps are changed. The clustering step can be parallelised over the pullback cover sets viajoblib
– though this can actually lower performance in small- and medium-size datasets. A range of pre-defined filter functions are also included, as well as covers in one and several dimensions, agglomerative clustering algorithms based on stopping rules to create flat cuts, and utilities for making transformers out of callables or out of other classes which have notransform
method.plot_static_mapper_graph
allows the user to visualise (in 2D or 3D) the Mapper graph arising from fit-transforming aMapperPipeline
to data, and offers a range of colouring options to correlate the graph’s structure with exogenous or endogenous information. It relies onplotly
for plotting and displaying metadata.plot_interactive_mapper_graph
adds interactivity to this, viaipywidgets
: specifically, the user can fine-tune some parameters involved in the definition of the Mapper pipeline, and observe in real time how the structure of the graph changes as a result. In this release, all hyperparameters involved in the covering and clustering steps are supported. The ability to fine-tune other hyperparameters will be considered for future versions.Added support for Python 3.8.
Bug Fixes¶
Fixed consistently incorrect documentation for the
fit_transform
methods. This has been achieved by introducing a class decoratoradapt_fit_transform_docs
which is defined in the newly introducedgtda.utils._docs.py
.
Backwards-Incompatible Changes¶
The library name change and the change in the name of the main module
giotto
are important major changes.There are now additional dependencies in the
python-igraph
,matplotlib
,plotly
, andipywidgets
libraries.
Thanks to our Contributors¶
This release contains contributions from many people:
Umberto Lupo, Lewis Tunstall, Guillaume Tauzin, Philipp Weiler, Julian Burella Pérez.
We are also grateful to all who filed issues or helped resolve them, asked and answered questions, and were part of inspiring discussions.
Release 0.1.3¶
Major Features and Improvements¶
None
Bug Fixes¶
Fixed a bug in
diagrams.Amplitude
causing the transformed array to be wrongly filled and added adequate test.
Backwards-Incompatible Changes¶
None.
Thanks to our Contributors¶
This release contains contributions from many people:
Umberto Lupo.
We are also grateful to all who filed issues or helped resolve them, asked and answered questions, and were part of inspiring discussions.
Major Features and Improvements¶
Added support for Python 3.5.
Bug Fixes¶
None.
Backwards-Incompatible Changes¶
None.
Thanks to our Contributors¶
This release contains contributions from many people:
Matteo Caorsi, Henry Tom (@henrytomsf), Guillaume Tauzin.
We are also grateful to all who filed issues or helped resolve them, asked and answered questions, and were part of inspiring discussions.
Release 0.1.1¶
Major Features and Improvements¶
Improved documentation.
Improved features of class
Labeller
.Improved features of class
PearsonDissimilarities
.Improved GitHub files.
Improved CI.
Bug Fixes¶
Fixed minor bugs from the first release.
Backwards-Incompatible Changes¶
The following class were renamed:
- class PearsonCorrelation
was renamed to class PearsonDissimilarities
Thanks to our Contributors¶
This release contains contributions from many people:
Umberto Lupo, Guillaume Tauzin, Matteo Caorsi, Olivier Morel.
We are also grateful to all who filed issues or helped resolve them, asked and answered questions, and were part of inspiring discussions.
Release 0.1.0¶
Major Features and Improvements¶
The following submodules where added:
giotto.homology
implements transformers to modify metric spaces or generate persistence diagrams.giotto.diagrams
implements transformers to preprocess persistence diagrams or extract features from them.giotto.time_series
implements transformers to preprocess time series or embed them in a higher dimensional space for persistent homology.giotto.graphs
implements transformers to create graphs or extract metric spaces from graphs.giotto.meta_transformers
implements conveniencegiotto.Pipeline
transformers for direct topological feature generation.giotto.utils
implements hyperparameters and input validation functions.giotto.base
implements aTransformerResamplerMixin
for transformers that have a resample method.giotto.pipeline
extends scikit-learn’s module by defining Pipelines that includeTransformerResamplers
.
Bug Fixes¶
None
Backwards-Incompatible Changes¶
None
Thanks to our Contributors¶
This release contains contributions from many people:
Guillaume Tauzin, Umberto Lupo, Philippe Nguyen, Matteo Caorsi, Julian Burella Pérez, Alessio Ghiraldello.
We are also grateful to all who filed issues or helped resolve them, asked and answered questions, and were part of inspiring discussions. In particular, we would like to thank Martino Milani, who worked on an early prototype of a Mapper implementation; although very different from the current one, it adopted an early form of caching to avoid recomputation in refitting, which was an inspiration for this implementation.
Release 0.1a.0¶
Initial release of the library, originally named giotto-learn
.