Problem of geometric field oversize

[MINE0126]

Hello, everyone,
I want to use Deepxde to compute a problem with a relatively sizeable geometric domain (300m x 100m). Many training points and more extended training costs are required without linear scaling of the geometric field. I have used linear scaling for the geometric domain question (scaled down by a factor of 100).
Is there any relevant paper that discusses whether it is correct and feasible to perform the linear scaling method in this way?
Thanks.

[lululxvi]

• https://en.wikipedia.org/wiki/Dimensional_analysis
• https://en.wikipedia.org/wiki/Change_of_variables