Table 2 Comparisons between the proposed algorithm and the state-of-the-art peer competitors
From: Automatic DNN architecture design using CPSOTJUTT for power system inspection
Model | CIFAR10 | CIFAR100 | #Parameters | GPU days | Categories |
|---|---|---|---|---|---|
DenseNet-40 (k = 12) | 5.24 | 24.42 | 1.0 M | – | MD |
ResNet-101 | 6.43 | 25.16 | 1.7 M | – | MD |
ResNet-1202 | 7.93 | 27.82 | 10.2 M | – | MD |
VGG | 6.66 | 28.05 | 20.01 M | – | MD |
Genetic CNN | 7.1 | 29.05 | – | 17 | SM |
Hierarchical Evolution | 3.63 | – | – | 300 | SM |
EAS | 4.23 | – | 23.4 M | 10 | SM |
Block-QGS-S | 4.38 | 20.65 | 6.1Â M | 90 | SM |
Large-scale Evolution | 5.4 | – | 5.4 M | 2750 | FA |
Large-scale Evolution | – | 23 | 40.3 | 2750 | FA |
CGP-CNN | 5.98 | – | 2.64 M | 27 | FA |
NAS | 6.01 | – | 2.5 M | 22,400 | FA |
AE-CNN | 4.3 | – | 2 M | 27 | FA |
AE-CNN | – | 20.85 | 5.4 M | 36 | FA |
CPGA-DNN | 3.76 | – | 1.56 M | 22 | FA |
CPGA-DNN | – | 17.35 | 4.67 | 34 | FA |