machine learning pattern classfication class SS2021
- prediction of valence and arousal
- how reliable / consistent are emotion annotations, how strongly do they agree / disagree?
- how is the agreement connected to major or minor features?
- how would you derive discrete quadrant-based classes from the raw annotatins?
- how would you deal with outliers in the annotations?
- how are the features distributed?
- are there pairs of features that are correlated, or features that seem like they will be obviously useful?