Table 2 The training time, measured in minutes, is given for both vanilla PINN and FE-PINN under different ratios and initial weight states.

From: Enhancing convergence speed with feature enforcing physics-informed neural networks using boundary conditions as prior knowledge

 

Ratio

FE-PINN

Vanilla PINN

\(\:\lambda\:=1\)

\(\:\lambda\:=1.2\)

\(\:\lambda\:=1.4\)

\(\:\lambda\:=1.6\)

\(\:\lambda\:=1.8\)

Case 1

20.72

1.2 + 21.5

Not Converged

Not Converged

Not Converged

Not Converged

Not Converged

Case 2

20.72

1.2 + 26.0

Not Converged

56.5

Not Converged

Not Converged

Not Converged

Case 3

20.72

1.3 + 24.1

Not Converged

Not Converged

54:6

57.8

53.6

Case 1

34.76

1.0 + 32.4

Not Converged

62.5

58.2

49.8

Not Converged

Case 2

34.76

1.3 + 18.2

30.0

Not Converged

Not Converged

Not Converged

Not Converged

Case 3

34.76

1.3 + 28.1

Not Converged

44.8

Not Converged

38.8

38.1

Case 1

38.28

1.1 + 24.0

29.1

47.8

Not Converged

Not Converged

48.3

Case 2

38.28

1.1 + 28.0

32.7

48.8

Not Converged

43.1

Not Converged

Case 3

38.28

1.1 + 17.3

Not Converged

52.3

Not Converged

Not Converged

47.4

Case 1

41.7

1.2 + 20.4

Not Converged

42.4

35.6

Not Converged

38.4

Case 2

41.7

1.2 + 21.1

Not Converged

56.6

42.3

57.6

Not Converged

Case 3

41.7

1.0 + 23.3

31.1

51.9

Not Converged

Not Converged

Not Converged

Case 1

43.81

1.3 + 26.5

Not Converged

Not Converged

41.7

44.9

Not Converged

Case 2

43.81

1.1 + 17.9

Not Converged

40.6

Not Converged

41.8

44.2

Case 3

43.81

1.2 + 22.0

Not Converged

51.4

37.2

47.2

39.8

Case 1

45.23

1.1 + 24.0

Not Converged

Not Converged

Not Converged

Not Converged

Not Converged

Case 2

45.23

1.3 + 18.0

26.3

38.3

Not Converged

Not Converged

Not Converged

Case 3

45.23

1.0 + 22.4

32.5

48.8

40.3

43.7

43.5