CHANGES

• [ENH] Clip outliers in FD and SPIKES group plots (#593)
• [ENH] Second revision of the classifier (#555):
  • Set matplotlib plugin to agg in docker image
  • Migrate scalings to sklearn pipelining system
  • Add Satra's feature selection for RFC (with thanks to S. Ghosh for his suggestion)
  • Make model selection compatible with sklearn Pipeline
  • Multiclass classification
  • Add feature selection filter based on Sites prediction (requires pinning to development sklearn-0.19)
  • Add RobustLeavePGroupsOut, replace RobustGridSearchCV with the standard GridSearchCV of sklearn.
  • Choice between RepeatedStratifiedKFold and RobustLeavePGroupsOut in mriqc_clf
  • Write cross-validation results to an .npz file.
• [ENH] First revision of the classifier (#553):
  • Add the possibility of changing the scorer function.
  • Unifize labels for raters in data tables (to rater_1)
  • Add the possibility of setting a custom decision threshold
  • Write the probabilities in the prediction file
  • Revised mriqc_clf processing flow
  • Revised labels file for ds030.
  • Add IQMs for ABIDE and DS030 calculated with MRIQC 0.9.6.
• ANNOUNCEMENT: Dropped support for Python<=3.4
• WARNING (#596):
  We have changed the default number of threads for ANTs. Using parallelism with ANTs
  causes numerical instability on the calculated measures. The most sensitive metrics to this
  problem are the kurtosis calculations on the intensities of regions and qi_2.