From: Features optimization selection in hidden layers of deep learning based on graph clustering
Classification accuracy | Features with 32 × 6 | Features with 16 × 6 | ||||
---|---|---|---|---|---|---|
KNN | ε−N | FC | KNN | ε−N | FC | |
Random | 0.545 | 0.459 | 0.496 | 0.556 | 0.460 | 0.487 |
SPEC | 0.698 | 0.721 | 0.774 | 0.658 | 0.701 | 0.739 |
ELasso | 0.773 | 0.759 | 0.782 | 0.693 | 0.734 | 0.755 |
LapCLasso | 0.744 | 0.801 | 0.795 | 0.711 | 0.803 | 0.768 |
Proposed with RC | 0.823 | 0.861 | 0.805 | 0.758 | 0.847 | 0.815 |
Proposed with NC | 0.788 | 0.750 | 0.840 | 0.749 | 0.796 | 0.812 |
Control group | KNN | \(\varepsilon - N\) | FC | |||
---|---|---|---|---|---|---|
Auto-encoder with features 64 × 6 | 0.788 | 0.761 | 0.830 |