Prediction-Based Maintenance of Existing Bridges Using Neural Network and Sensitivity Analysis

Table 4

Configuration test results.

Parameters	Parameters	value
		Training	Validation
Number of hidden layers	1, 2	0.5489	0.1154
Number of hidden neurons	5, 10	0.4236	0.8454
Learning rate	0.01, 0.001; SGD, RMSProp, Adagrad,	0.0003^a	0.0038^a
Learning algorithm	Adadelta, Adam, AdaMax, Nadam	5.513 × 10^−14a	8.490 × 10^−15a
Hidden activation function	ReLu, tanh	0.4904	0.7622

^aSignificant at 95% level of confidence.