半监督学习与混合GAN架构的深入研究
半监督学习参数分析
不同数据集分类结果
在半监督学习(Semi-unsupervised Learning,SuSL)中,对不同数据集进行了分类实验,以下是相关结果:
| 数据集 | 模型 | % 标记数据 (准确率 ± 标准差)% | | | |
| — | — | — | — | — | — |
| | | 0 (UL) | 20 (SuSL) | 20 (SSL) | 100 (SL) |
| MNIST | GM - DGM | 90.0 ± 1.5 | 92.5 ± 0.3 | 97.7 ± 0.1 | 98.7 ± 0.1 |
| | GM - DGM+ | 96.1 ± 0.3 | 99.0 ± 0.1 | 99.4 ± 0.0 | 99.5 ± 0.0 |
| FMNIST | GM - DGM | 75.8 ± 0.3 | 78.2 ± 1.1 | 86.9 ± 0.1 | 90.0 ± 0.1 |
| | GM - DGM+ | 74.3 ± 0.3 | 81.7 ± 0.4 | 89.4 ± 0.3 | 92.0 ± 0.3 |
| KMNIST | GM - DGM | 78.1 ± 0.5 | 88.5 ± 0.5 | 89.4 ± 0.3 | 94.5 ± 0.2 |
| | GM - DGM+ | 76.7 ± 0.5 | 91.3 ± 0.5 | 96.0 ± 0.3 | 97.6 ± 0.1 |
| K49MNIST | GM - DGM | 49.9 ± 1.1 | 69.0 ± 0.7 | 67.4 ± 0.7