Research Article

Convolution Neural Network-Based Sensitive Security Parameter Identification and Analysis

Table 3

Hyperparameters of CNN models.

Hyperparameter

Filter size{32, 64, 128, 256}{4, 8, 16, 32}
Epoch2015
Output size (number of classes)54
Batch size100100
Kernel size
Kernel initializerhe_normalhe_normal
Conv strides
Pooling strides22
Output activation typeSoftmaxSoftmax
Optimizerrmsproprmsprop
Loss functioncategorical_crossentropycategorical_crossentropy
PaddingSameSame
Pooing sizes after each conv layer