Table 6 Hyperparameter sensitivity analysis results on \(R^2\).

From: physically interpretable residual strength prediction of corroded pipelines via symbolic Bayesian networks

Learning rate

Adagrad

RMSprop

Adam

SGD

Adamax

NAdam

0.001

0.145

0.956

0.959

0.927

0.844

0.965

0.005

0.614

0.942

0.959

0.878

0.904

0.909

0.01

0.722

0.911

0.948

0.866

0.943

0.900

Back to article page