Skip to main content

Table 6 PINNs approximation errors along with train and test losses for different training points

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

Training sample

\(L_{2}\)

\(L_{\infty }\)

RMSE

Train loss

Test loss

100

8.455e−05

1.788e−04

8.104e−05

2.30e−08

2.60e−08

200

3.828e−05

7.238e−05

2.823e−05

6.73e−09

8.65e−09

300

1.169e−05

3.098e−05

1.092e−05

2.17e−09

2.50e−09

400

1.183e−05

2.837e−05

1.129e−05

1.74e−09

1.92e−09

500

6.215e−05

8.342e−05

6.472e−06

9.58e−09

1.01e−08

600

1.294e−05

3.697e−05

1.153e−05

1.50e−09

1.49e−09

700

3.745e−05

6.052e−05

3.746e−06

8.76e−09

9.69e−09

800

2.418e−05

5.974e−05

2.411e−06

1.85e−09

1.81e−09

900

2.405e−05

5.712e−05

2.405e−06

2.87e−09

3.14e−09

1000

2.124e−05

4.905e−05

1.852e−05

6.57e−09

6.38e−09

1100

7.426e−05

8.561e−05

7.426e−06

9.24e−10

9.00e−10

1200

4.253e−05

9.261e−05

4.049e−05

3.11e−09

2.78e−09

1300

3.224e−05

4.919e−05

3.226e−06

8.35e−09

8.38e−09

1400

1.520e−05

3.334e−05

1.520e−06

2.04e−09

1.93e−09

1500

5.734e−05

1.622e−04

7.655e−06

2.19e−09

2.14e−09

1600

1.332e−05

2.000e−05

1.332e−06

3.92e−09

4.22e−09

1700

4.861e−06

1.469e−05

4.861e−07

5.22e−09

5.27e−09

1800

9.205e−06

7.795e−06

1.097e−06

5.86e−09

6.01e−09

1900

6.040e−06

2.390e−05

2.7848e−06

1.09e−09

1.07e−09

2000

6.114e−07

1.526e−06

5.976e−07

4.12e−10

4.17e−10