Skip to main content

Table 5 PINNs approximation errors along with train and test losses for different learning rates

From: A deep learning approach: physics-informed neural networks for solving a nonlinear telegraph equation with different boundary conditions

Activation functions

\(L_{2}\)

\(L_{\infty }\)

RMSE

Train loss

Test loss

1e-01

1.637e−02

7.103e−02

1.577e−02

2.85e−09

8.82e−05

1e−02

2.58e−02

2.867e−02

2.867e−03

1.76e−06

1.66e−06

1e−03

8.028e−05

2.110e−04

7.01e−05

2.33e−10

2.39e−10

2e−03

1.463e−04

1.753e−04

1.435e−04

2.50e−09

2.76e−09

3e−03

3.023e−03

3.050e−03

1.371e−04

2.91e−09

2.94e−09

4e−03

5.542e−03

5.562e−03

5.541e−03

1.04e−07

1.27e−07

5e−03

8.303e−03

8.316e−03

8.315e−04

3.45e−08

2.73e−08

6e−03

3.32e−03

3.41e−03

3.067e−03

2.97e−09

2.98e−09

7e−03

3.921e−03

3.940e−03

3.920e−03

7.53e−08

7.41e−08

8e−03

1.992e−02

1.994e−02

1.992e−03

2.92e−09

2.92e−09

9e−03

2.223e−02

2.231e−02

2.222e−02

8.17e−07

1.23e−06

1e−04

6.158e−04

1.414e−03

6.115e−04

1.04e−07

1.27e−07

1e−05

5.687e−04

1.849e−03

5.752e−04

2.95e−07

3.62e−07

1e−06

5.270e−04

1.734e−03

5.249e−04

3.80e−07

4.78e−07

1e−07

2.778e−01

3.0192e−01

2.778e−02

2.95e−07

3.62e−07