Application of Artificial Intelligence for the Estimation of Concrete and Reinforcement Consumption in the Construction of Integral Bridges

Table 4

Activation functions of a multilayer perceptron model of the artificial neural network.

Function	Mark	Explanation	Range
Identity	x	Only in the output layer
Rectified linear unit function		Neuron activation is forwarded directly as an output if positive, and if negative, it is forwarded to 0. It has been shown to have 6 times better convergence than a hyperbolic tangent function
A hyperbolic tangent		Neuron activation is forwarded directly as an output if positive, and if negative, it is forwarded to 0. It has been shown to have 6 times better convergence than a hyperbolic tangent function