- Raw FFT results into features
- Saving and loading npz properly
- Some tests (unsuccessfull)
- Refactoring of the notebook (deleted extra cells, added md-titles, cleared output)
- mat_to_pd()
- FFT histogram bins as features (PCA shows that it's a bad set of features + RF shows results ~0.5 roc_auc)
- save_to_npz()
- t-test showed that number of peaks in data does not matter