Version 0.9.0 (2026-08-24)#
Upcoming breaking changes for v1.0#
The next release of MNE-Connectivity will be v1.0, which will include breaking changes. If you do not want the v1.0 release to break your code, you should pin your MNE-Connectivity version to <=0.9 until v1.0 is released and you can update your code to accommodate any required changes.
The following breaking changes are anticipated:
Separation of functions accepting MNE data objects and array-like inputs
Requiring spectral representations of data to be provided for spectral connectivity analysis functions
Changes to improve consistency of all-to-all vs. symmetric indexing of connectivity
Requiring
mne.Infoobjects as input to connectivity containersRemoving VAR-specific methods (
eigvals,predict,simulate, etc) from non-VAR connectivity containersRemoving epoch-specific methods (
append,combine) from non-epoched connectivity containersMoving spectral Granger causality methods from
spectral_connectivity_epochs()andspectral_connectivity_time()to dedicated functions
For more details and the most up-to-date information, see the following GitHub issue: Expected changes for 1.0 #440.
Bugfixes#
Fix bug where the
get_data()method of connectivity containers discarded the imaginary part of complex-valued data whenoutput="dense", by Thomas Binns. (#412)Fix bug where connectivity in
make_signals_in_freq_bands()could be simulated outside of the epoch duration, by Thomas Binns. (#418)Fix bug where the
predict()andsimulate()methods of connectivity containers could not be used with VAR models with multiple lags, by Thomas Binns. (#420)Fix a bug in
make_surrogate_evoked_data()where sampled cutpoints were interpreted as numbers of sections instead of split indices, by Rebecca Stevenson. (#447)
API changes by deprecation#
Drop
fskipparameter fromspectral_connectivity_epochs()andspectral_connectivity_time()in favour offdecim, by Thomas Binns. (#392)Deprecate
sfreqparameter default value of2*np.piinphase_slope_index()in favour ofNone, which must be specified when input data is an array-like, by Thomas Binns. (#409)Begin deprecation of
make_surrogate_datain favour ofmake_surrogate_resting_data(), by Thomas Binns. (#415)Begin deprecation of
n_timesparameter ofmake_signals_in_freq_bands()in favour ofdurationas the way to specify epoch length. Also begins deprecation ofconnection_delayto be specified in units of seconds rather than samples, by Thomas Binns. (#418)
New features#
Add
imcohas a method tospectral_connectivity_time(), by Thomas Binns. (#404)Add
make_surrogate_evoked_data()function for generating surrogates from evoked data, and add support formne.Epochsdata inmake_surrogate_resting_data(), by Thomas Binns. (#415)Add
connection_timeparameter tomake_signals_in_freq_bands(), enabling connectivity to be simulated in a specific time range, by Thomas Binns. (#418)Allow epoched data as input to the
predict()method of connectivity containers storing a non-dynamic VAR model, by Thomas Binns. (#420)