Version 0.13 (2016-09-27)

Changelog

• Add new class AcqParserFIF to parse Elekta/Neuromag MEG acquisition info, allowing e.g. collecting epochs according to acquisition-defined averaging categories by Jussi Nurminen

• Adds automatic determination of FIR filter parameters filter_length, l_trans_bandwidth, and h_trans_bandwidth and adds phase argument in e.g. in mne.io.Raw.filter() by Eric Larson

• Adds faster n_fft='auto' option to mne.io.Raw.apply_hilbert() by Eric Larson

• Adds new function mne.time_frequency.csd_array to compute the cross-spectral density of multivariate signals stored in an array, by Nick Foti

• Add order params ‘selection’ and ‘position’ for mne.viz.plot_raw() to allow plotting of specific brain regions by Jaakko Leppakangas

• Added the ability to decimate mne.Evoked objects with mne.Evoked.decimate() by Eric Larson

• Add generic array-filtering function mne.filter.filter_data() by Eric Larson

• mne.viz.plot_trans now also shows head position indicators by Christian Brodbeck

• Add label center of mass function mne.Label.center_of_mass() by Eric Larson

• Added mne.viz.plot_ica_properties() that allows plotting of independent component properties similar to pop_prop in EEGLAB. Also mne.preprocessing.ICA has mne.preprocessing.ICA.plot_properties() method now. Added by Mikołaj Magnuski

• Add second-order sections (instead of (b, a) form) IIR filtering for reduced numerical error by Eric Larson

• Add interactive colormap option to image plotting functions by Jaakko Leppakangas

• Add support for the University of Maryland KIT system by Christian Brodbeck

• Add support for *.elp and *.hsp files to the KIT2FIFF converter and mne.channels.read_dig_montage by Teon Brooks and Christian Brodbeck

• Add option to preview events in the KIT2FIFF GUI by Christian Brodbeck

• Add approximation of size of io.Raw, Epochs, and Evoked in repr() by Eric Larson

• Add possibility to select a subset of sensors by lasso selector to mne.viz.plot_sensors() and mne.viz.plot_raw() when using order=’selection’ or order=’position’ by Jaakko Leppakangas

• Add the option to plot brain surfaces and source spaces to viz.plot_bem() by Christian Brodbeck

• Add the --filterchpi option to mne browse_raw, by Felix Raimundo

• Add the --no-decimate option to mne make_scalp_surfaces to skip the high-resolution surface decimation step, by Eric Larson

• Add new class mne.decoding.EMS to transform epochs with the event-matched spatial filters and add ‘cv’ parameter to mne.decoding.compute_ems(), by Jean-Remi King

• Added mne.time_frequency.EpochsTFR and average parameter in mne.time_frequency.tfr_morlet() and mne.time_frequency.tfr_multitaper() to compute time-frequency transforms on single trial epochs without averaging, by Jean-Remi King and Alex Gramfort

• Added mne.decoding.TimeFrequency to transform signals in scikit-learn pipelines, by Jean-Remi King

• Added mne.decoding.UnsupervisedSpatialFilter providing interface for scikit-learn decomposition algorithms to be used with MNE data, by Jean-Remi King and Asish Panda

• Added support for multiclass decoding in mne.decoding.CSP, by Jean-Remi King and Alexandre Barachant

• Components obtained from mne.preprocessing.ICA are now sorted by explained variance, by Mikołaj Magnuski

• Adding an EEG reference channel using mne.add_reference_channels() will now use its digitized location from the FIFF file, if present, by Chris Bailey

• Added interactivity to mne.preprocessing.ICA.plot_components() - passing an instance of io.Raw or Epochs in inst argument allows to open component properties by clicking on component topomaps, by Mikołaj Magnuski

• Adds new function mne.viz.plot_compare_evokeds() to show multiple evoked time courses at a single location, or the mean over a ROI, or the GFP, automatically averaging and calculating a CI if multiple subjects are given, by Jona Sassenhagen

• Added transform_into parameter into mne.decoding.CSP to retrieve the average power of each source or the time course of each source, by Jean-Remi King

• Added support for reading MaxShield (IAS) evoked data (e.g., from the acquisition machine) in mne.read_evokeds() by Eric Larson

• Added support for functional near-infrared spectroscopy (fNIRS) channels by Jaakko Leppakangas

• Added mne.io.Raw.acqparser convenience attribute for mne.AcqParserFIF by Eric Larson

• Added example of Representational Similarity Analysis, by Jean-Remi King

BUG

• Fixed a bug where selecting epochs using hierarchical event IDs (HIDs) was and-like instead of or-like. When doing e.g. epochs[('Auditory', 'Left')], previously all trials that contain 'Auditory' and 'Left' (like 'Auditory/Left') would be selected, but now any conditions matching 'Auditory' or 'Left' will be selected (like 'Auditory/Left', 'Auditory/Right', and 'Visual/Left'). This is now consistent with how epoch selection was done without HID tags, e.g. epochs[['a', 'b']] would select all epochs of type 'a' and type 'b'. By Eric Larson

• Fixed Infomax/Extended Infomax when the user provides an initial weights matrix by Jair Montoya Martinez

• Fixed the default raw FIF writing buffer size to be 1 second instead of 10 seconds by Eric Larson

• Fixed channel selection order when MEG channels do not come first in mne.preprocessing.maxwell_filter() by Eric Larson

• Fixed color ranges to correspond to the colorbar when plotting several time instances with mne.viz.plot_evoked_topomap() by Jaakko Leppakangas

• Added units to mne.io.read_raw_brainvision() for reading non-data channels and enable default behavior of inferring channel type by unit by Jaakko Leppakangas and Pablo-Arias

• Fixed minor bugs with mne.Epochs.resample() and mne.Epochs.decimate() by Eric Larson

• Fixed a bug where duplicate vertices were not strictly checked by mne.simulation.simulate_stc() by Eric Larson

• Fixed a bug where some FIF files could not be read with mne.io.show_fiff() by Christian Brodbeck and Eric Larson

• Fixed a bug where merge_grads=True causes mne.viz.plot_evoked_topo() to fail when plotting a list of evokeds by Jaakko Leppakangas

• Fixed a bug when setting multiple bipolar references with set_bipolar_reference() by Marijn van Vliet.

• Fixed image scaling in mne.viz.plot_epochs_image() when plotting more than one channel by Jaakko Leppakangas

• Fixed mne.preprocessing.Xdawn to fit shuffled epochs by Jean-Remi King

• Fixed a bug with channel order determination that could lead to an AssertionError when using mne.Covariance matrices by Eric Larson

• Fixed the check for CTF gradient compensation in mne.preprocessing.maxwell_filter() by Eric Larson

• Fixed the import of EDF files with encoding characters in mne.io.read_raw_edf() by Guillaume Dumas

• Fixed mne.Epochs to ensure that detrend parameter is not a boolean by Jean-Remi King

• Fixed bug with mne.realtime.FieldTripClient.get_data_as_epoch when picks=None which crashed the function by Mainak Jas

• Fixed reading of units in .elc montage files (from UnitsPosition field) so that mne.channels.Montage objects are now returned with the pos attribute correctly in meters, by Chris Mullins

• Fixed reading of BrainVision files by Phillip Alday:

• Greater support for BVA files, especially older ones: alternate text coding schemes with fallback to Latin-1 as well as units in column headers

• Use online software filter information when present

• Fix comparisons of filter settings for determining “strictest”/”weakest” filter

• Weakest filter is now used for heterogeneous channel filter settings, leading to more consistent behavior with filtering methods applied to a subset of channels (e.g. Raw.filter with picks != None).

• Fixed plotting and timing of Annotations and restricted addition of annotations outside data range to prevent problems with cropping and concatenating data by Jaakko Leppakangas

• Fixed ICA plotting functions to refer to IC index instead of component number by Andreas Hojlund and Jaakko Leppakangas

• Fixed bug with picks when interpolating MEG channels by Mainak Jas.

• Fixed bug in padding of Stockwell transform for signal of length a power of 2 by Johannes Niediek

API

• The add_eeg_ref argument in core functions like mne.io.read_raw_fif() and mne.Epochs has been deprecated in favor of using mne.set_eeg_reference() and equivalent instance methods like raw.set_eeg_reference(). In functions like mne.io.read_raw_fif() where the default in 0.13 and older versions is add_eeg_ref=True, the default will change to add_eeg_ref=False in 0.14, and the argument will be removed in 0.15.

• Multiple aspects of FIR filtering in MNE-Python has been refactored:

  1. New recommended defaults for l_trans_bandwidth='auto', h_trans_bandwidth='auto', and filter_length='auto'. This should generally reduce filter artifacts at the expense of slight decrease in effective filter stop-band attenuation. For details see Defaults in MNE-Python. The default values of l_trans_bandwidth=h_trans_bandwidth=0.5 and filter_length='10s' will change to 'auto' in 0.14.

  2. The filter_length=None option (i.e. use len(x)) has been deprecated.

  3. An improved phase='zero' zero-phase FIR filtering has been added. Instead of running the designed filter forward and backward, the filter is applied once and we compensate for the linear phase of the filter. The previous phase='zero-double' default will change to phase='zero' in 0.14.

  4. A warning is provided when the filter is longer than the signal of interest, as this is unlikely to produce desired results.

  5. Previously, if the filter was as long or longer than the signal of interest, direct FFT-based computations were used. Now a single code path (overlap-add filtering) is used for all FIR filters. This could cause minor changes in how short signals are filtered.

• Support for Python 2.6 has been dropped, and the minimum supported dependencies are NumPy 1.8, SciPy 0.12, and Matplotlib 1.3 by Eric Larson

• When CTF gradient compensation is applied to raw data, it is no longer reverted on save of mne.io.Raw.save() by Eric Larson

• Adds mne.time_frequency.csd_epochs to replace mne.time_frequency.csd_compute_epochs for naming consistency. mne.time_frequency.csd_compute_epochs is now deprecated and will be removed in mne 0.14, by Nick Foti

• Weighted addition and subtraction of Evoked as ev1 + ev2 and ev1 - ev2 have been deprecated, use explicit mne.combine_evoked(..., weights='nave') instead by Eric Larson

• Deprecated support for passing a list of filenames to mne.io.Raw constructor, use mne.io.read_raw_fif() and mne.concatenate_raws() instead by Eric Larson

• Added options for setting data and date formats manually in mne.io.read_raw_cnt() by Jaakko Leppakangas

• Now channels with units of ‘C’, ‘µS’, ‘uS’, ‘ARU’ and ‘S’ will be turned to misc by default in mne.io.read_raw_brainvision() by Jaakko Leppakangas

• Add mne.io.anonymize_info() function to anonymize measurements and add methods to mne.io.Raw, mne.Epochs and mne.Evoked, by Jean-Remi King

• Now it is possible to plot only a subselection of channels in mne.viz.plot_raw() by using an array for order parameter by Jaakko Leppakangas

• EOG channels can now be included when calling mne.preprocessing.ICA.fit() and a proper error is raised when trying to include unsupported channels by Alexander Rudiuk

• mne.concatenate_epochs() and mne.compute_covariance() now check to see if all Epochs instances have the same MEG-to-Head transformation, and errors by default if they do not by Eric Larson

• Added option to pass a list of axes to mne.viz.plot_epochs_image() by Mikołaj Magnuski

• Constructing IIR filters in mne.filter.construct_iir_filter() defaults to output='ba' in 0.13 but this will be changed to output='sos' by Eric Larson

• Add zorder parameter to mne.Evoked.plot() and derived functions to sort allow sorting channels by e.g. standard deviation, by Jona Sassenhagen

• The baseline parameter of mne.Epochs.apply_baseline() is set by default (None, 0), by Felix Raimundo

• Adds mne.Evoked.apply_baseline() to be consistent with mne.Epochs.apply_baseline(), by Felix Raimundo

• Deprecated the baseline parameter in mne.Evoked, by Felix Raimundo

• The API of mne.SourceEstimate.plot() and mne.viz.plot_source_estimates() has been updated to reflect current PySurfer 0.6 API. The config_opts parameter is now deprecated and will be removed in mne 0.14, and the default representation for time will change from ms to s in mne 0.14. By Christian Brodbeck

• The default dataset location has been changed from examples/ in the MNE-Python root directory to ~/mne_data in the user’s home directory, by Eric Larson

• A new option set_env has been added to mne.set_config() that defaults to False in 0.13 but will change to True in 0.14, by Eric Larson

• The compensation parameter in mne.io.read_raw_fif() has been deprecated in favor of the method mne.io.Raw.apply_gradient_compensation() by Eric Larson

• mne.decoding.EpochsVectorizer has been deprecated in favor of mne.decoding.Vectorizer by Asish Panda

• The epochs_data parameter has been deprecated in mne.decoding.CSP, in favour of the X parameter to comply to scikit-learn API, by Jean-Remi King

• Deprecated mne.time_frequency.cwt_morlet and mne.time_frequency.single_trial_power in favour of mne.time_frequency.tfr_morlet() with parameter average=False, by Jean-Remi King and Alex Gramfort

• Add argument mask_type to mne.read_events() and mne.find_events() to support MNE-C style of trigger masking by Teon Brooks and Eric Larson

• Extended Infomax is now the new default in mne.preprocessing.infomax() (extended=True), by Clemens Brunner

• mne.io.read_raw_eeglab() and mne.read_epochs_eeglab() now take additional argument uint16_codec that allows to define the encoding of character arrays in set file. This helps in rare cases when reading a set file fails with TypeError: buffer is too small for requested array. By Mikołaj Magnuski

• Added mne.decoding.TemporalFilter to filter data in scikit-learn pipelines, by Asish Panda

• mne.preprocessing.create_ecg_epochs() now includes all the channels when picks=None by Jaakko Leppakangas

• mne.set_eeg_reference() now allows moving from a custom to an average EEG reference by Marijn van Vliet

Authors

The committer list for this release is the following (sorted by alphabetical order):

• Alexander Rudiuk

• Alexandre Barachant

• Alexandre Gramfort

• Asish Panda

• Camilo Lamus

• Chris Holdgraf

• Christian Brodbeck

• Christopher J. Bailey

• Christopher Mullins

• Clemens Brunner

• Denis A. Engemann

• Eric Larson

• Federico Raimondo

• Félix Raimundo

• Guillaume Dumas

• Jaakko Leppakangas

• Jair Montoya

• Jean-Remi King

• Johannes Niediek

• Jona Sassenhagen

• Jussi Nurminen

• Keith Doelling

• Mainak Jas

• Marijn van Vliet

• Michael Krause

• Mikolaj Magnuski

• Nick Foti

• Phillip Alday

• Simon-Shlomo Poil

• Teon Brooks

• Yaroslav Halchenko