Research Article
Domain Adaption Based on ELM Autoencoder
| (a) Recognition accuracy with DA using a SVM Classifier (Office dataset + Caltech-256) |
| | Algorithm | | | | | | |
| | DA-SA1 [7] | 44.30% | 36.80% | 32.90% | 36.80% | 29.60% | 24.90% | | DA-SA2 [7] | 44.50% | 38.60% | 34.20% | 37.30% | 31.60% | 28.40% | | PCA [7] | 46.10% | 41.05% | 39.30% | 39.20% | 35.00% | 31.80% | | GFK [16] | 44.82% | 37.91% | 37.10% | 38.37% | 31.42% | 29.14% | | SA-ELM-DA(SURF, sigmoid) | 46.62% | 41.15% | 39.02% | 40.88% | 35.93% | 33.17% | | SA-ELM-DA(SURF, linear) | 27.82% | 34.00% | 36.04% | 23.17% | 32.03% | 31.00% | | SA-ELM-DA(CNN, linear) | 61.62% | 41.29% | 52.73% | 52.78% | 35.87% | 45.13% | | SA-ELM-DA(CNN, sigmoid) | 80.80% | 47.92% | 68.89% | 67.31% | 36.79% | 52.60% |
|
|
| (b) Recognition accuracy with DA using a SVM Classifier (Office dataset + Caltech-256) |
| | Algorithm | | | | | | |
| | DA-SA1 [7] | 36.10% | 38.90% | 73.60% | 42.50% | 34.60% | 75.40% | | DA-SA2 [7] | 32.50% | 35.30% | 73.60% | 37.30% | 34.20% | 80.50% | | PCA [7] | 38.80% | 39.40% | 77.90% | 39.60% | 38.90% | 82.30% | | GFK [16] | 37.92% | 36.14% | 74.60% | 39.81% | 34.93% | 79.10% | | SA-ELM-DA(SURF, sigmoid) | 39.76% | 41.46% | 80.76% | 41.05% | 39.10% | 83.88% | | SA-ELM-DA(SURF, linear) | 36.04% | 37.11% | 82.32% | 36.00% | 33.24% | 76.44% | | SA-ELM-DA(CNN, linear) | 46.34% | 34.97% | 43.28% | 60.21% | 57.34% | 46.79% | | SA-ELM-DA(CNN, sigmoid) | 49.43% | 45.03% | 52.61% | 72.88% | 64.34% | 55.34% |
|
|
