Research Article
Fully Kernected Neural Networks
Table 2
Validation accuracy of CNN, KNN, and our FKNN on CIFAR10.
| Methods | Network-depth | Validation accuracy (%) | Improvement on CNN |
| CNN | ResNet-18 | 91.18 | 0 | CNN | ResNet-110 | 93.57 | 0 | KNN | ResNet-110 | 95.31 | 1.74% | FKNN (ours) | ResNet-18 | 95.73 | 4.55% |
| CNN | DenseNet-40 | 94.76 | 0 | KNN | DenseNet-40 | 94.92 | 0.16% | FKNN (ours) | DenseNet-40 | 96.45 | 1.69% |
| CNN | GoogLeNet-22 | 94.95 | 0 | KNN | GoogLeNet-22 | 94.84 | ā0.11% | FKNN (ours) | GoogLeNet-22 | 95.58 | 0.74% |
|
|
We mark the best results in bold values.
|