Table 4 Average accuracy and top-1/top-5 error results on test set.
From: Enhancing deep neural network training efficiency and performance through linear prediction
Model | Epochs (Accuracy) | Optimal Model | ||||
|---|---|---|---|---|---|---|
10 | 20 | 30 | Accuracy | Top-1 error | Top-5 error | |
Vgg16(DEMON) | 37.91% | 50.34% | 56.83% | 57.38% | 0.4447 | 0.1632 |
Vgg16(PLP) | 37.88% | 50.80% | 56.79% | 57.31% | 0.4317 | 0.1617 |
Vgg16(SGD) | 36.15% | 49.33% | 55.51% | 56.42% | 0.4511 | 0.1654 |
Resnet18(DEMON) | 38.01% | 50.97% | 57.10% | 59.88% | 0.4290 | 0.1563 |
Resnet18(PLP) | 37.82% | 50.75% | 56.98% | 60.01% | 0.4312 | 0.1603 |
Resnet18(SGD) | 35.62% | 49.86% | 56.16% | 58.76% | 0.4384 | 0.1640 |
GoogLeNet(DEMON) | 37.31% | 51.45% | 56.13% | 59.86% | 0.4387 | 0.1642 |
GoogLeNet(PLP) | 37.51% | 51.03% | 56.67% | 58.37% | 0.4333 | 0.1606 |
GoogLeNet(SGD) | 34.95% | 49.65% | 55.79% | 57.70% | 0.4421 | 0.1664 |
