Table 1 Training time, in minutes, for different \(\:\lambda\:\) to converge to a loss value of 10− 3 in case 1 and case 2.

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

 

\(\:\lambda\:=0.1\)

\(\:\lambda\:=0.5\)

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

\(\:\lambda\:=1.1\)

\(\:\lambda\:=1.5\)

\(\:\lambda\:=5\)

Case 1

Not Converged

40.883

17.900

15.283

25.050

14.266

Case 2

Not Converged

13.450

27.550

25.583

15.933

28.616

Back to article page