What’s new in the package

A catalog of new features, improvements, and bug-fixes in each release.

v0.12.dev

• Deprecate covariances_X and cospectrum. #442 by @qbarthelemy

• Add Python Array API support for NumPy/PyTorch backend transparency in core utility modules (base, covariance, distance, mean, geodesic, tangentspace, ajd, kernel, median), enabling execution on both NumPy arrays and PyTorch tensors with optional GPU acceleration and autograd support. #433 by @bruAristimunha

• Move geometry modules (ajd, base, covariance, distance, geodesic, kernel, mean, median, tangentspace, test) from pyriemann.utils to a new standalone pyriemann.geometry subpackage. The old import paths (e.g. pyriemann.utils.mean, pyriemann.utils.kernel, pyriemann.utils.test) still work as backward-compatibility shims but emit a DeprecationWarning; rename to pyriemann.geometry.<module>. The private pyriemann.utils._backend and pyriemann._helpers modules also moved into pyriemann.geometry so that the subpackage is fully standalone (no internal pyriemann imports outside of itself). Module pyriemann.utils.utils is renamed to pyriemann.utils._check with the same shim+warning. Tests for moved modules are renamed test_utils_* → test_geometry_*. #445 by @bruAristimunha

• Enhance pyriemann.geometry.geodesic.geodesic() to accept alpha as an ndarray of shape (...,), allowing a different geodesic position per stacked matrix pair. #396 by @Fashad-Ahmed

• Add example on Riemannian curvature analysis of sentence trajectories in language model embeddings, demonstrating how local metric tensors (SPD matrices) capture geometric structure in LLM latent spaces and enable classification of semantically distinct sentences using MDM. #448 by @SzczepanK112 and @gcattan

• Add example on simulated SPD matrices to compare metrics. #451 by @qbarthelemy

• Add Bini-Meini-Poloni (BMP) mean pyriemann.geometry.mean.mean_bmp(), and Cheap mean pyriemann.geometry.mean.mean_cheap(). #449 by @qbarthelemy

v0.11 (April 2026)

• Enhance pyriemann.datasets.sample_gaussian_spd() adding support for dispersion defined as a covariance matrix. #412 by @thibaultdesurrel

• Add pyriemann.clustering.GaussianMixture. #411 by @qbarthelemy

• Enhance pyriemann.classification.NearestConvexHull to support “euclid” metric. #415 by @qbarthelemy

• Deprecate mean_covariance, renamed into pyriemann.geometry.mean.gmean(). #419 by @qbarthelemy

• Correct log-Euclidean parallel transport. #420 by @qbarthelemy

• Add CI caching for Zenodo datasets to speed up documentation builds and avoid rate limiting. #417 by @bruAristimunha

• Enhance pyriemann.classification.TSClassifier and pyriemann.transfer.TLClassifier to support classifiers without sample weights. #422 by @qbarthelemy

• Enhance pyriemann.clustering.PotatoField, allowing a different metric per potato and adding a parameter method_combination. #423 by @qbarthelemy

• Modernize documentation: migrate from Bootstrap to Furo theme, add card-based API navigation with sphinx-design, and fix Sphinx build warnings. #424 by @bruAristimunha

• Add ORCID identifiers and new contributors to CITATION.cff. #424 by @bruAristimunha

• Add functions to compute inner products. #428 by @qbarthelemy

• Add broadcast compatibility for utility functions, enabling batched SPD/HPD matrix operations. Includes broadcast-compatible Mahalanobis distance. #426 by @bruAristimunha

• Complete example on artifact detection by Riemannian potato field, adding a metric by potato. #431 by @DavoudYneuro

• Enhance pyriemann.geometry.covariance.covariance_scm() to estimate weighted sample covariance matrices. #434 by @qbarthelemy

v0.10 (January 2026)

• Add example in gallery to compare clustering algorithms on synthetic datasets. #374 by @qbarthelemy

• Enhance pyriemann.classification.MeanField, to be used as a feature extractor. #377 by @qbarthelemy

• Update pyRiemann from Python 3.9 - 3.11 to 3.10 - 3.12. #378 by @qbarthelemy

• Deprecate fit_transform() of pyriemann.spatialfilters.AJDC due to incompatible dimensions. #382 by @qbarthelemy

• Add pyriemann.datasets.RandomOverSampler for data augmentation of positive-definite matrices. #387 by @qbarthelemy

• Add pyriemann.geometry.tangentspace.transport() for parallel transport of positive-definite matrices. #388 by @qbarthelemy

• Enhance pyriemann.datasets.make_matrices() to generate invertible, orthogonal, unitary or non-square matrices. #389 by @qbarthelemy

• Deprecate mean_identity. #392 by @qbarthelemy

• Enhance pyriemann.tangentspace.TangentSpace to support HPD matrices. #394 by @qbarthelemy

• Speedup of the ALM mean pyriemann.geometry.mean.mean_alm(). #398 by @qbarthelemy

• Add pyriemann.geometry.geodesic.geodesic_chol() and pyriemann.geometry.mean.mean_chol(). #399 by @qbarthelemy

• Add pyriemann.geometry.tangentspace.transport_logchol() for parallel transport with log-Cholesky metric. #400 by @qbarthelemy

• Add Thompson metric: pyriemann.geometry.distance.distance_thompson(), pyriemann.geometry.geodesic.geodesic_thompson(), and pyriemann.geometry.mean.mean_thompson(). #401 by @qbarthelemy

• Fix sklearn error This Pipeline instance is not fitted yet. #406 by @qbarthelemy

• Add pyriemann.classification.NearestConvexHull. #405 by @qbarthelemy

v0.9 (July 2025)

• Add conjugate transpose operator pyriemann.geometry.base.ctranspose() for real- and complex-valued ndarrays. #348 by @qbarthelemy

• Add pyriemann.embedding.TSNE, a Riemannian t-SNE implementation and update example comparing embeddings. #347 by @thibaultdesurrel

• Fix matplotlib warning. #351 by @qbarthelemy

• Enhance pyriemann.geometry.mean.mean_power() using init, tol and maxiter parameters when p=0. #353 by @toncho11

• Enhance pyriemann.geometry.distance.pairwise_distance() to support HPD matrices for “euclid”, “harmonic”, “logchol” and “logeuclid” metrics. #350 by @qbarthelemy

• Avoid duplicating code when using joblib. #359 by @qbarthelemy

• Fix sklearn warning This Pipeline instance is not fitted yet. #358 by @qbarthelemy

• Add pyriemann.geometry.tangentspace.log_map() and pyriemann.geometry.tangentspace.exp_map(). #363 by @qbarthelemy

• Add pyriemann.clustering.MeanShift, a Riemannian mean shift clustering algorithm. #364 by @qbarthelemy

• Fix pyriemann.spatialfilters.Xdawn to remove the parameters sample_weight that is not used on the fit_transformation. #371 by @bruAristimunha

v0.8 (February 2025)

• Enhance pyriemann.geometry.mean.mean_ale() adding init parameter, and add function check_init() useful to all ajd and mean functions allowing initialization. #328 by @qbarthelemy

• Enhance mean_covariance to support “power” and “poweuclid” metrics. #329 by @qbarthelemy

• Add tangent space alignment (TSA) in transfer learning module. TLStretch is deprecated and renamed into pyriemann.transfer._estimators.TLScale, and TSclassifier into pyriemann.classification.TSClassifier. #320 by @qbarthelemy

• Add an example using fNIRS data with a new estimator called HybridBlocks for classifying HbO and HbR signals. #323 by @timnaher

• Add directional derivatives pyriemann.geometry.base.ddexpm() and pyriemann.geometry.base.ddlogm(), and correct pyriemann.geometry.tangentspace.log_map_logeuclid() and pyriemann.geometry.tangentspace.exp_map_logeuclid(). #332 by @gabelstein

• Add pyriemann.geometry.tangentspace.exp_map_wasserstein(), pyriemann.geometry.tangentspace.log_map_wasserstein() and pyriemann.geometry.geodesic.geodesic_wasserstein(). #331 by @gabelstein

• Enhance pyriemann.datasets.make_matrices(), to generate symmetric and Hermitian matrices, and add parameters defining the normal distribution to draw eigen vectors. Deprecate generate_random_spd_matrix. #339 by @qbarthelemy

• Enhance TSA, adding weights to transformers, and generalizing pyriemann.transfer._estimators.TLRotate from one-to-one to many-to-one domain adaptation in tangent space. #337 by @qbarthelemy

• Enhance pyriemann.geometry.kernel.kernel_euclid() and pyriemann.geometry.kernel.kernel_logeuclid() adding Cref parameter, and correct pyriemann.geometry.kernel.kernel_riemann() when Y is different from X and Cref is None. #340 by @qbarthelemy

• Speedup of the Wasserstein mean pyriemann.geometry.mean.mean_wasserstein() and extension of transport function pyriemann.geometry.tangentspace.transport(). #341 by @gabelstein

v0.7 (October 2024)

• Add kernel parameter to pyriemann.embedding.LocallyLinearEmbedding. #293 by @gabelstein

• Add possibility for target_domain parameter of pyriemann.transfer._estimators.TLCenter to be empty, forcing transform() to recenter matrices to the last fitted domain. #292 by @brunaafl

• Enhance pyriemann.geometry.ajd.ajd_pham() and pyriemann.geometry.mean.mean_ale() functions to process HPD matrices. #299 by @qbarthelemy

• Add partial_fit function to pyriemann.preprocessing.Whitening for online applications. #277 by @qbarthelemy and @brentgaisford

• Add get_weights fixture to conftest and complete tests. #305 by @qbarthelemy

• Enhance pyriemann.estimation.Shrinkage to process HPD matrices. #307 by @qbarthelemy

• Add remote sensing examples on radar image clustering. #306 by @AmmarMian

• Add sample_weight parameter to MDM.fit_predict, Potato.fit, Potato.partial_fit, PotatoField.fit, PotatoField.partial_fit, Whitening.partial_fit. #309 by @qbarthelemy

• Update pyRiemann from Python 3.8 - 3.10 to 3.9 - 3.11. #310 by @qbarthelemy

• Add pyriemann.geometry.distance.distance_logchol() to compute log-Cholesky distance. #311 by @qbarthelemy

• Add ajd_method parameter to pyriemann.spatialfilters.CSP. #313 by @qbarthelemy

• Add pyriemann.geometry.distance.distance_poweuclid() and pyriemann.geometry.mean.mean_poweuclid() to use power Euclidean metric. #312 by @qbarthelemy

• Enhance pyriemann.geometry.mean.mean_power(): add init parameter and fix default initialization provided in the associated paper. #324 by @toncho11

• Add pyriemann.geometry.distance.distance_chol(), pyriemann.geometry.geodesic.geodesic_logchol(), pyriemann.geometry.mean.mean_logchol(), and correct pyriemann.geometry.distance.distance_logchol(). #322 by @gabelstein

v0.6 (April 2024)

• Update pyRiemann from Python 3.7 - 3.9 to 3.8 - 3.10. #254 by @qbarthelemy

• Speedup pairwise distance function pyriemann.geometry.distance.pairwise_distance() by adding individual functions for ‘euclid’, ‘harmonic’, ‘logeuclid’ and ‘riemann’ metrics. #256 by @gabelstein

• Add pyriemann.geometry.test.is_real_type() to check the type of input arrays and add pyriemann.geometry.covariance.covariance_scm() allowing to process complex-valued inputs for ‘scm’ covariance estimator. #251 by @qbarthelemy

• Update to Read the Docs v2. #260 by @qbarthelemy

• Correct pyriemann.geometry.distance.distance_wasserstein() and pyriemann.geometry.distance.distance_kullback(), keeping only real part. #267 by @qbarthelemy

• Deprecate input covmats for mean functions, renamed into X. #252 by @qbarthelemy

• Add support for complex covariance estimation for ‘lwf’, ‘mcd’, ‘oas’ and ‘sch’ estimators. #274 by @gabelstein

• Deprecate input covtest for predict of pyriemann.classification.KNearestNeighbor, renamed into X. #259 by @qbarthelemy

• Correct check for kernel_fct param of pyriemann.classification.SVC. #272 by @qbarthelemy

• Add sample_weight parameter in TLCenter, TLStretch and TLRotate. #273 by @apmellot

• Deprecate HankelCovariances, renamed into pyriemann.estimation.TimeDelayCovariances. #275 by @qbarthelemy

• Add an example on augmented covariance matrix. #276 by @carraraig

• Remove function make_covariances. #280 by @qbarthelemy

• Speedup pyriemann.estimation.TimeDelayCovariances. #281 by @qbarthelemy

• Enhance ajd module and add a generic pyriemann.geometry.ajd.ajd() function. #238 by @qbarthelemy

• Add pyriemann.utils.viz.plot_bihist(), pyriemann.utils.viz.plot_biscatter() and pyriemann.utils.viz.plot_cov_ellipse() for display. #287 by @qbarthelemy and @gcattan

• Add pyriemann.estimation.CrossSpectra, and deprecate CospCovariances renamed into pyriemann.estimation.CoSpectra. #288 by @qbarthelemy

v0.5 (Jun 2023)

• Fix pyriemann.geometry.distance.pairwise_distance() for non-symmetric metrics. #229 by @qbarthelemy

• Fix mean_covariance used with keyword arguments. #230 by @qbarthelemy

• Add functions to test HPD and HPSD matrices, pyriemann.geometry.test.is_herm_pos_def() and pyriemann.geometry.test.is_herm_pos_semi_def(). #231 by @qbarthelemy

• Add function pyriemann.datasets.make_matrices() to generate SPD, SPSD, HPD and HPSD matrices. Deprecate function pyriemann.datasets.make_covariances. #232 by @qbarthelemy

• Add tests for matrix operators and distances for HPD matrices, complete doc and add references. #234 by @qbarthelemy

• Enhance tangent space module to process HPD matrices. #236 by @qbarthelemy

• Fix regression introduced in pyriemann.spatialfilters.Xdawn() by #214. #242 by @qbarthelemy

• Fix pyriemann.geometry.kernel.kernel_euclid() applied on non-symmetric matrices. #245 by @qbarthelemy

• Add argument squared to all distances. #246 by @qbarthelemy

• Correct transform and predict_proba of pyriemann.classification.MeanField. #247 by @qbarthelemy

• Enhance mean module to process HPD matrices. #243 by @qbarthelemy

• Correct pyriemann.geometry.distance.distance_mahalanobis(), keeping only real part. #249 by @qbarthelemy

• Fix pyriemann.datasets.sample_gaussian_spd() used with sampling_method=rejection on 2D matrices. #250 by @mhurte

v0.4 (Feb 2023)

• Add exponential and logarithmic maps for three main metrics: ‘euclid’, ‘logeuclid’ and ‘riemann’. pyriemann.geometry.tangentspace.tangent_space() is splitted in two steps: (i) log_map_*() projecting SPD matrices into tangent space depending on the metric; and (ii) pyriemann.geometry.tangentspace.upper() taking the upper triangular part of matrices. Similarly, pyriemann.geometry.tangentspace.untangent_space() is splitted into (i) pyriemann.geometry.tangentspace.unupper() and (ii) exp_map_*(). The different metrics for tangent space mapping can now be defined into pyriemann.tangentspace.TangentSpace, then used for transform() as well as for inverse_transform(). #195 by @qbarthelemy

• Enhance AJD: add init to pyriemann.geometry.ajd.ajd_pham() and pyriemann.geometry.ajd.rjd(), add warm_restart to pyriemann.spatialfilters.AJDC. #196 by @qbarthelemy

• Add parameter sampling_method to pyriemann.datasets.sample_gaussian_spd(), with rejection accelerating 2x2 matrices generation. #198 by @Artim436

• Add geometric medians for Euclidean and Riemannian metrics: pyriemann.geometry.median_euclid(), and pyriemann.geometry.median_riemann(), and add an example in gallery to compare means and medians on synthetic datasets. #200 by @qbarthelemy

• Add score() to pyriemann.regression.KNearestNeighborRegressor. #205 by @qbarthelemy

• Add Transfer Learning module and examples, including RPA and MDWM. #189 by @plcrodrigues, @qbarthelemy and @sylvchev

• Add class distinctiveness function to measure the distinctiveness between classes on the manifold, pyriemann.classification.class_distinctiveness(), and complete an example in gallery to show how it works on synthetic datasets. #215 by @MSYamamoto

• Add example on ensemble learning applied to functional connectivity, and add pyriemann.geometry.base.nearest_sym_pos_def(). #202 by @mccorsi and @sylvchev

• Add kernel matrices representation pyriemann.estimation.Kernels and complete example comparing estimators. #217 by @qbarthelemy

• Add a new covariance estimator, robust fixed point covariance, and add kwds arguments for all covariance based functions and classes. #220 by @qbarthelemy

• Add example in gallery on frequency band selection using class distinctiveness measure. #219 by @MSYamamoto

• Add pyriemann.geometry.covariance.covariance_mest() supporting three robust M-estimators (Huber, Student-t and Tyler) and available for all covariance based functions and classes; and add an example on robust covariance estimation for corrupted data. Add also pyriemann.geometry.distance.distance_mahalanobis() between between vectors and a Gaussian distribution. #223 by @qbarthelemy

v0.3 (July 2022)

• Correct spectral estimation in pyriemann.geometry.covariance.cross_spectrum() to obtain equivalence with SciPy. #131 by @qbarthelemy

• Add instantaneous, lagged and imaginary coherences in pyriemann.geometry.covariance.coherence(), and pyriemann.estimation.Coherences. #132 by @qbarthelemy

• Add partial_fit in pyriemann.clustering.Potato, useful for an online update; and update example on artifact detection. #133 by @qbarthelemy

• Deprecate pyriemann.utils.viz.plot_confusion_matrix as sklearn integrate its own version. #135 by @sylvchev

• Add Ando-Li-Mathias (ALM) mean in pyriemann.geometry.mean.mean_alm(). #56 by @sylvchev

• Add Schaefer-Strimmer covariance estimator in pyriemann.geometry.covariance.covariances(), and an example to compare estimators. #59 by @sylvchev

• Refactor tests + fix refit of pyriemann.tangentspace.TangentSpace. #136 by @sylvchev

• Add pyriemann.clustering.PotatoField, and an example on artifact detection. #142 by @qbarthelemy

• Add sampling SPD matrices from a Riemannian Gaussian distribution in pyriemann.datasets.sample_gaussian_spd(). #140 by @plcrodrigues

• Add new function pyriemann.datasets.make_gaussian_blobs() for generating random datasets with SPD matrices. #140 by @plcrodrigues

• Add module pyriemann.utils.viz in API, add pyriemann.utils.viz.plot_waveforms(), and add an example on ERP visualization. #144 by @qbarthelemy

• Add a special form covariance matrix pyriemann.geometry.covariance.covariances_X(). #147 by @qbarthelemy

• Add masked and NaN means with Riemannian metric: pyriemann.geometry.mean.maskedmean_riemann() and pyriemann.geometry.mean.nanmean_riemann(). #149 by @qbarthelemy and @sylvchev

• Add corr option in pyriemann.geometry.covariance.normalize(), to normalize covariance into correlation matrices. #153 by @qbarthelemy

• Add block covariance matrix: pyriemann.estimation.BlockCovariances and pyriemann.geometry.covariance.block_covariances(). #154 by @gabelstein

• Add Riemannian Locally Linear Embedding: pyriemann.embedding.LocallyLinearEmbedding and pyriemann.embedding.locally_linear_embedding(). #159 by @gabelstein

• Add Riemannian Kernel Function: pyriemann.geometry.kernel.kernel_riemann(). #159 by @gabelstein

• Fix fit in pyriemann.channelselection.ElectrodeSelection. #166 by @qbarthelemy

• Add power mean estimation in pyriemann.geometry.mean.mean_power(). #170 by @qbarthelemy and @plcrodrigues

• Add example in gallery to compare classifiers on synthetic datasets. #175 by @qbarthelemy

• Add predict_proba in pyriemann.classification.KNearestNeighbor, and correct attribute classes_. #171 by @qbarthelemy

• Add Riemannian Support Vector Machine classifier: pyriemann.classification.SVC. #175 by @gabelstein and @qbarthelemy

• Add Riemannian Support Vector Machine regressor: pyriemann.regression.SVR. #175 by @gabelstein and @qbarthelemy

• Add K-Nearest-Neighbor regressor: pyriemann.regression.KNearestNeighborRegressor. #164 by @gabelstein, @qbarthelemy and @agramfort

• Add Minimum Distance to Mean Field classifier: pyriemann.classification.MeanField. #172 by @qbarthelemy and @plcrodrigues

• Add example on principal geodesic analysis (PGA) for SSVEP classification. #169 by @qbarthelemy

• Add pyriemann.geometry.distance.distance_harmonic(), and sort functions by their names in code, doc and tests. #183 by @qbarthelemy

• Parallelize functions for dataset generation: pyriemann.datasets.make_gaussian_blobs(). #179 by @sylvchev

• Fix dispersion when generating datasets: pyriemann.datasets.sample_gaussian_spd(). #179 by @sylvchev

• Enhance base and distance functions, to process ndarrays of SPD matrices. #186 and #187 by @qbarthelemy

• Enhance utils functions, to process ndarrays of SPD matrices. #190 by @qbarthelemy

• Enhance means functions, with faster implementations and warning when convergence is not reached. #188 by @qbarthelemy

v0.2.7 (June 2021)

• Add example on SSVEP classification. #81 by @sylvchev

• Fix compatibility with scikit-learn v0.24

• Correct probas of pyriemann.classification.MDM. #100 by @qbarthelemy

• Add predict_proba for pyriemann.clustering.Potato, and an example on artifact detection. #105 by @qbarthelemy

• Add a normalization function for covariance matrices, pyriemann.geometry.covariance.normalize(). #111 by @qbarthelemy

• Add weights to Pham’s AJD algorithm pyriemann.geometry.ajd.ajd_pham(). #112 by @qbarthelemy

• Add pyriemann.geometry.covariance.cross_spectrum(), fix pyriemann.geometry.covariance.cospectrum(); pyriemann.geometry.covariance.coherence() output is kept unchanged. #118 by @qbarthelemy

• Add a function to compute non-diagonality weights of matrices, pyriemann.geometry.covariance.get_nondiag_weight(). #119 by @qbarthelemy

• Add pyriemann.spatialfilters.AJDC for BSS and gBSS, with an example on artifact correction. #120 by @qbarthelemy

• Add pyriemann.preprocessing.Whitening, with optional dimension reduction. #122 by @qbarthelemy

v0.2.6 (March 2020)

• Remove support for Python 2, and update code for better scikit-learn v0.22 support. #79 by @alexandrebarachant

v0.2.5 (January 2018)

• Add BilinearFilter.

• Add a permutation test for generic scikit-learn estimator.

• Enhance stats module, with distance based t-test and f-test.

• Remove two way permutation test.

• Add FlatChannelRemover. #30 by @kingjr

• Add support for Python 3.5 in travis.

• Add Shrinkage transformer. #38 by @alexandrebarachant

• Add Coherences transformer.

• Add Embedding class and pyriemann.geometry.distance.pairwise_distance(). #54 by @plcrodrigues

v0.2.4 (June 2016)

• Improve documentation.

• Add TSclassifier for out-of the box tangent space classification.

• Add Wasserstein distance and mean.

• Add KNearestNeighbor classifier.

• Add softmax probabilities for MDM.

• Add CSP for covariance matrices.

• Add approximate joint diagonalization algorithms: JADE, PHAM, UWEDGE.

• Add ALE mean.

• Add multiclass CSP.

• Correct param name in CospCovariances to comply to scikit-learn.

• Correct attributes name in most modules to comply to the scikit-learn naming convention.

• Add HankelCovariances estimation.

• Add SPoC spatial filtering.

• Add harmonic mean.

• Add Kullback-Leibler mean.

v0.2.3 (November 2015)

• Add multiprocessing for MDM with joblib.

• Add Kullback-Leibler divergence.

• Add Riemannian Potato.

• Add sample_weight for mean estimation and MDM.

v0.2.2 (June 2015)

• Add possibility to use a dictionary to define metrics used for MDM.

• Add svd argument in ERPCovariances.

v0.1 (April 2015)

• Add MDM and first utils (distance, mean, geodesic, covariance).

• Add FgMDM, TangentSpace, FGDA.

• Add ElectrodeSelection, Covariances, ERPCovariances, XdawnCovariances, Xdawn.

• Add examples for motor imagery ad ERP.