Table 1.

The list of hyperparameters in CNNs to be tuned

Hyperparameter	Range
Kernel size for convolution	3, 7, 15, 20, 30, 40, 50
Number of kernels (in two convolution layers)	6:13, 13:26, 19:38, 32:64, 45:90, 64:128, 128:256
Pooling method	Max pooling, average pooling,
Pooling in second layer	Global max pooling, local max pooling
Number of units in hidden layer (ratio to input layer)	1/3, 1/2, 2/3, 3/4, 1
Learning algorithm	Adam, AdaGrad, AdaDelta, Momentum SGD