Extended feature naming to all univariate functions. If no specific function is implemented for a given univariate function and that the output has the same shape as the input on the feature dimension, then a generic list of feature names is passed. Possibility to rename these feature by passing ch_names to mne_features.extract_features(). By Paul ROUJANSKY in #60.
Added root mean square (RMS) and percentile univariate feature extraction functions. By Hubert Banville in #61.
Added log and ratios_triu to mne_features.univariate.compute_pow_freq_bands() to allow computing log-power and log-ratios of power and controlling whether all possible band ratios should be computed. By Hubert Banville in #62.
Extended feature naming to all bivariate functions, with the same option to pass ch_names to mne_features.extract_features() as for univariate functions. By Hubert Banville in #63.
Fixed the behavior of the edge parameter in mne_features.univariate.compute_spect_edge_freq(). Valid values for this parameter are now None or a list of float between 0 and 1 (percentages). By Jean-Baptiste SCHIRATTI in #52.
Fixed channel name replacement with overlapping channel names (e.g., Cz and FCz) which affected mne_features.extract_features() with return_as_df=True and when ch_names are provided. By Hubert Banville in #63.
Fixed feature_extraction.FeatureFunctionTransformer’s check_input method following sickit-learn 1.0 release: passing reset=True. By Paul ROUJANSKY in #70.