v.0.2.2 (February 2019)

• Added Cython to doc requirements
• Removed multiprocessing from doc requirements
• Fixed hanging comment in Nifti docstring
• Added Dockerfile
• Support predictions for single electrode locations

v0.2.1 (July 2018)

Minor bugfix release that suppresses unnecessary warnings thrown when supereeg is imported

v0.2.0 (July 2018)

Note: this version breaks previous compatibility with model objects.

Changes:

• Models are now computed in log space to improve stability of the computations
• Models now (internally) maintain a "numerator" and "denominator" field, which allows finer grained control over how information is integrated across models
• Added additional parameters to Model objects:
  • mo.rbf_width (default : 20) which is used as the RBF width for spatial smoothing and interpolation
  • meta information now includes a stability property to indicate whether any operations that might affect model stability have been performed. A warning is outputted and some operations may not complete if stability is False.
• Added/modified methods to Model objects:
  • Support for basic algebra with Model objects. E.g. m3 = m1 + m2 yields a combined Model object m3 that incorporates information from both the m1 and m2 Model objects. This is equivalent to m3 = Model([m1, m2]). Similarly, m1_recon = m3 - m2 yields a model that is functionally equivalent to m1 (i.e. m1_recon is based on model m3, but factors out m2). This provides a fast way to remove a single model from a combined Model object without redoing all of the internal calculations; e.g. m1_recon will be equivalent (with respect to how it reconstructs data) to m1. Note: the subtraction operation yields unstable Model objects that no longer retains per-model weights via the denominator property. (These models may still be used to predict and reconstruct data, but they cannot be expanded to incorporate data from additional locations.)
  • Model.get_locs() that returns the locations in the model.
  • Added z_transform flag to Model.get_model(). If z_transform==False, the Pearson correlation matrix is returned. If z_transform==True, the z-transformed correlation matrix is returned instead.
  • Added a Model.get_slice() function that returns the indexed model object at a set of desired locations.
• The n_secs attribute of Brain objects has been renamed to dur.
• Enhanced the Brain.plot_locs() method. Observed locations are now plotted in blue, filtered locations are plotted in cyan, and reconstructed locations are plotted in red.
• Miscellaneous bug fixes (see code for details)
• Overall code cleanup

v0.1.1 (April 2018)

Enhancements:

• added 'pyFR_k20_r10_6mm' model to load function
• modified plot function for very large models

Bugs fixed:

• fixed bug where license could not be found when installing from pip