Version 0.8 (2014-06-25)#



  • Fix incorrect times attribute when stc was computed using apply_inverse after decimation at epochs stage for certain, arbitrary sample frequencies by Denis Engemann

  • Fix corner case error for step-down-in-jumps permutation test (when step-down threshold was high enough to include all clusters) by Eric Larson

  • Fix selection of total number of components via float when picking ICA sources by Denis Engemann and Qunxi Dong

  • Fix writing and reading transforms after modification in measurement info by Denis Engemann and Martin Luessi and Eric Larson

  • Fix pre-whitening / rescaling when estimating ICA on multiple channels without covariance by Denis Engemann

  • Fix ICA pre-whitening, avoid recomputation when applying ICA to new data by Denis Engemann


  • The minimum numpy version has been increased to 1.6 from 1.4.

  • Epochs object now has a selection attribute to track provenance of selected Epochs. The length of the drop_log attribute is now the same as the length of the original events passed to Epochs. In earlier versions it had the length of the events filtered by event_id. Epochs has also now a plot_drop_log method.

  • Deprecate Epochs.drop_picks in favor of a new method called drop_channels

  • Deprecate labels_from_parc and parc_from_labels in favor of read_labels_from_annot and write_labels_to_annot

  • The default of the new add_dist option of setup_source_space to add patch information will change from False to True in MNE-Python 0.9

  • Deprecate read_evoked and write_evoked in favor of read_evokeds and write_evokeds. read_evokeds will return all Evoked instances in a file by default.

  • Deprecate setno in favor of condition in the initialization of an Evoked instance. This affects mne.fiff.Evoked and read_evokeds, but not read_evoked.

  • Deprecate mne.fiff module, use instead e.g. instead of mne.fiff.Raw.

  • Pick functions (e.g., pick_types) are now in the mne namespace (e.g. use mne.pick_types).

  • Deprecated ICA methods specific to one container type. Use, ICA.get_sources ICA.apply and ICA.plot_* for processing Raw, Epochs and Evoked objects.

  • The default smoothing method for mne.stc_to_label will change in v0.9, and the old method is deprecated.

  • As default, for ICA the maximum number of PCA components equals the number of channels passed. The number of PCA components used to reconstruct the sensor space signals now defaults to the maximum number of PCA components estimated.


The committer list for this release is the following (preceded by number of commits):

  • 418 Denis A. Engemann

  • 284 Alexandre Gramfort

  • 242 Eric Larson

  • 155 Christian Brodbeck

  • 144 Mainak Jas

  • 49 Martin Billinger

  • 49 Andrew Dykstra

  • 44 Tal Linzen

  • 37 Dan G. Wakeman

  • 36 Martin Luessi

  • 26 Teon Brooks

  • 20 Cathy Nangini

  • 15 Hari Bharadwaj

  • 15 Roman Goj

  • 10 Ross Maddox

  • 9 Marmaduke Woodman

  • 8 Praveen Sripad

  • 8 Tanay

  • 8 Roan LaPlante

  • 5 Saket Choudhary

  • 4 Nick Ward

  • 4 Mads Jensen

  • 3 Olaf Hauk

  • 3 Brad Buran

  • 2 Daniel Strohmeier

  • 2 Federico Raimondo

  • 2 Alan Leggitt

  • 1 Jean-Remi King

  • 1 Matti Hämäläinen