Table 3. Multi-Layer Perceptron (MLP) and convolutional neural network methods: Hyperparameters’ domain and the corresponding tuned values at the data sets under consideration.

The N_n, N_e, lr, in respect, represents the number of neurons of the hidden layer, the number of epochs, and the learning rate. The drop shows the maximum dropout in the max pooling layer.

method / data set	parameters
	N n	N e	lr	activation	optimiser	Drop
MLP	{2, 3, …, 200}	[10, 50000]	[1e-6, 1e-2]	{Identity, Logistic, Tanh, ReLu}	{LBFGS, SGD, ADAM},	[0.1, 0.8]
MLP at Demo	173	31270	-	Identity	LBFGS	-
MLP at Fixation	190	58325	-	ReLu	LBFGS	-
CNN at Fixation	192	100	0.0001	ReLu	ADAM	0.1
MLP at Demo-Fixation	158	49150	-	Tanh	LBFGS	-
CNN at Demo-Fixation	192	200	0.0001	ReLu	ADAM	0.1
MLP at IA	34	66253	-	Logistic	LBFGS
MLP at Demo-IA	50	13953	-	Logistic	LBFGS