基于相关熵和流形正则化的图像聚类
|
|
时照群, 刘兆伟, 刘惊雷
|
Image clustering based on correntropy and manifold regularization
|
|
Zhaoqun Shi, Zhaowei Liu, Jinglei Liu
|
|
表3 本文算法和七个对比算法在五个数据集上的聚类结果：精度、归一化互信息和纯度
|
Table 3 Clustering results of ACC，NMI and PUR by our algorithm and other algorithms on five image datasets
|
|
| Dataset | ACC |
|---|
| k⁃means | NMF | GNMF | CGNMF | GDNMF | GLoRSS | l1⁃CNMF | CRNMF |
|---|
| CMU PIE | 24.23% | 51.32% | 76.22% | 76.36% | 76.07% | 72.35% | 73.89% | 78.30% | | ORL | 49.02% | 58.66% | 60.15% | 60.69% | 56.07% | 56.31% | 61.62% | 61.73% | | UMIST | 40.05% | 49.34% | 53.70% | 54.67% | 54.47% | 51.31% | 54.35% | 63.13% | | YaleB | 07.25% | 30.47% | 31.33% | 34.47% | 33.39% | 33.07% | 33.97% | 35.42% | | COIL20 | 59.72% | 65.77% | 71.84% | 75.46% | 79.83% | 73.12% | 76.33% | 81.67% | | Dataset | NMI | | k⁃means | NMF | GNMF | CGNMF | GDNMF | GLoRSS | l1⁃CNMF | CRNMF | | CMU PIE | 53.65% | 73.26% | 88.31% | 88.57% | 88.13% | 85.14% | 87.25% | 92.52% | | ORL | 69.23% | 50.14% | 73.38% | 74.76% | 73.46% | 73.76% | 73.88% | 74.93% | | UMIST | 58.04% | 46.92% | 70.84% | 74.61% | 71.07% | 70.91% | 74.13% | 76.75% | | YaleB | 08.18% | 38.12% | 42.70% | 42.19% | 42.08% | 41.29% | 42.78% | 53.06% | | COIL20 | 70.05% | 71.98% | 79.53% | 85.99% | 88.33% | 82.30% | 87.06% | 90.03% | | Dataset | PUR | | k⁃means | NMF | GNMF | CGNMF | GDNMF | GLoRSS | l1⁃CNMF | CRNMF | | CMU PIE | 30.62% | 59.49% | 80.67% | 86.25% | 87.46% | 84.33% | 86.70% | 86.79% | | ORL | 59.20% | 60.96% | 64.22% | 66.15% | 61.97% | 62.23% | 65.41% | 66.18% | | UMIST | 45.73% | 54.73% | 66.83% | 67.84% | 73.04% | 64.19% | 67.68% | 67.65% | | YaleB | 12.21% | 32.15% | 38.91% | 39.98% | 39.99% | 38.34% | 39.57% | 40.49% | | COIL20 | 68.59% | 74.57% | 73.97% | 84.77% | 85.83% | 83.31% | 85.43% | 86.39% |
|
|
|