Table 1 Accuracy per layer and statistical features of their filters for VGG-16 trained on CIFAR-100.

From: Towards a universal mechanism for successful deep learning

VGG-16 on CIFAR-100

Layer

\({N}_{f}\)

\({F}_{s}\)

\(F{C}_{s}\)

Accuracy

\(n\)

\({N}_{c}\)

\({C}_{s}\)

13

512

1 × 1

512

0.745

277.1

1.7

7.7

10

512

2 × 2

2048

0.752

16.3

2.6

2.0

7

256

4 × 4

4096

0.577

117.9

2.8

2.7

4

128

8 × 8

8192

0.439

552.4

5.1

2.8

2

64

16 × 16

16,384

0.352

987.8

5.8

3.2

  1. \({N}_{f}\) number of filters of layers terminating with max-pooling, \({F}_{s}\) filter sizes, \(F{C}_{S}\) size of trained FC layer connected to the output units, \(n\) average noise per filter, \({N}_{c}\) average number of clusters per filter, \({C}_{S}\) average cluster size.
Back to article page