- [feature] AutoFeat: Automatic training feature selection mechanism, with option to define a subselection of channels and frequencies for this selection.
- [feature] "Combination training" mechanism.
- [feature] Online BCI scenario generation, from a training attempt. Generates a readily usable OpenViBE scenario with selected features & classifier trained.
- [feature][experimental] Added "Advanced Mode" in the main GUI, enabling replaying signal files with a trained classifier (only for PSD and Connectivity pipelines)
- [feature] ERD/ERS available for Connectivity pipeline
- [dependencies] Changed from PyQt5 to PySide2
- [dependencies] Imposed some versions (MNE, matplotlib) due to modified behaviors (or removed/refactored functionailites), to be corrected in later versions of HF.
- [pipeline][experimental] One-class MI classification pipeline
- [feature][bugfix] Mechanism for handling invalid (NaN) values after extraction. Trials containing NaN values are automatically discarded at the visualization step. Extracted files are still usable in the visualisation part (hence no more "empty display"), but with a warning for the user.
- [test] Basic tests (& github action automation), more to come!
- [bugfix] Work thread management (template training scenarios, loading viz files when using multiple metrics)
- Various other [bugfix]'s...