model accuracy and robustness

[realamirhe]

The PD data has been collected for 3 test on each patient so model prediction for slice of data may causes inconsistency in report

PD-Classification/autoencoder.py

Line 127 in 793f2a1

y_pred = np.round(model.predict(x_test))

Majority vote can make model evaluation more robust but may affect accuracy score.

---

Also normal fold selection may cause a user be both in train and test set in each fold evaluation
which can be prevented by filtering on unique patient identifiers