Research Article
Hilbert–Schmidt Independence Criterion Regularization Kernel Framework on Symmetric Positive Definite Manifolds
Table 3
Classification accuracy (%) of different kernels.
| | Method | COIL-20 | ETH80 |
| | KPCA | 81.05 | 72.61 | | logEuc-SC | 89.76 | 73.75 | | RSR | 93.95 | 77.74 | | TSC | 80.16 | 65.88 | | Riem-DLSC | 87.74 | 75.63 | | KLRM-DL | 97.33 | 80.13 | | HSIC-SL | 96.87 | 82.40 | | RLPP | 85.89 | 74.08 | | RLPP-HR (log-linear) | 87.58 | 77.88 | | RLPP-HR (log-Gaussian) | 88.55 | 77.25 | | RLPP-HR (LogDet divergence) | 87.58 | 74.38 | | CDL | 94.54 | 79.92 | | CDL-HR (log-linear) | 97.58 | 82.63 | | CDL-HR (log-Gaussian) | 97.26 | 81.38 | | CDL-HR (LogDet divergence) | 97.18 | 82.5 | | KSLR | 96.24 | 81.66 | | KSLR-HR (log-linear) | 97.98 | 85 | | KSLR-HR (log-Gaussian) | 98.87 | 84.5 | | KSLR-HR (LogDet divergence) | 97.82 | 84.88 | | HRGDA | 98.87 | 84.88 |
|
|
