Table 2 Effects of hyperparameters on the performance of a multilayer perceptron neural network for predicting bread texture (activation function, regularization, and dropout)

		Train set		Test set
		R2	RMSE	R2	RMSE
		Activation function	Sigmoid	0.3224±0.0999	0.1918±0.0141	0.4005±0.1497	0.1942±0.0245
Tanh	0.7508±0.0029		0.1166±0.0007	0.6965±0.0107	0.1392±0.0024
Linear	0.7496±0.0019		0.1169±0.0004	0.6934±0.0107	0.1399±0.0024
ELU	0.7550±0.0040		0.1156±0.0009	0.6938±0.0073	0.1398±0.0017
ReLU	0.9882±0.0112		0.0229±0.0108	0.7504±0.0471	0.1257±0.0119
Leaky_ReLU	0.9684±0.0130		0.0404±0.0097	0.7846±0.0303	0.1170±0.0081
Regularization		0.9981±0.0021	0.0088±0.0049	0.7891±0.0402	0.1156±0.0107
Regularization+Dropout		0.9667±0.0108	0.0421±0.0069	0.8109±0.0272	0.1096±0.0079