Shengfeng
Shengfeng
Well, it's essentially the same when testing the inverse problem
Hi, Thanks for reading my code. I just updated the file, it should be working now. The "test.py" is written mainly out of the needs of debugging so it is...
the loss include both data loss and equation loss, if both the data loss and equation loss are declined around 1e-5, the flow field should be well fitted (according to...
Well, this looks like your data is wrong. You can check if your data is non-dimensional data, since the underlying equations presented in the code is non-dimensional NS equations. PIus,...
你数值计算的结果可以通过升阻力系数，st数自行判定，关于你选择的区域大小问题，一般来讲不要太小就行（至少包含一些网格，不要在网格尺度以下），太小会出一些拟合问题。 具体你跑出来不收敛可能是其他的问题，你自己可以再检查一下，这个代码虽然比较老了，但是正确性是有保证的。 祝好， 盛峰 At 2024-02-27 22:12:06, "HCwinwinwin" ***@***.***> wrote: image.png (view on web) Thanks！ Before inputting the data, I processed it according to the steps mentioned, and it seems...
Well, if you are using this code, you could just cite this repository. If you are using hyperparameters similar to those in my paper, you can cite my paper. It...
您好 我在代码里面有个validation func的部分，会输出训练过程中的相对误差，你可以检查一下，如果这个相对误差的值没有有效下降的话那就是训练的batch和epoch的选择可能不合理，如果这个相对误差的值在0.1以下，那就是画图的时候数据的行和列有问题，初学者容易在行优先和列优先等问题上出问题，可以仔细检查一下 Regards, Shengfeng 在 2025-06-03 16:55:27，"chendiandian421" ***@***.***> 写道： chendiandian421 created an issue (Shengfeng233/PINN-for-NS-equation#6) comparep.gif (view on web) compareu.gif (view on web) comparev.gif (view on web) 我是一个初学者，导师想要我用Pinn来预测其他的东西，我看您提供的代码与数据库很完整，所以想着通过复现你的来进行学习。跑出来的动图显示预测值与实际值相差很大，请您指导下可能的原因是？感谢! — Reply...
我看很多人都有这个问题，可能是依赖库的版本不一样导致的，我有空复现一下这个问题 在 2025-11-21 09:47:00，"zzl1213880" ***@***.***> 写道： zzl1213880 left a comment (Shengfeng233/PINN-for-NS-equation#6) 作者大大，我也用到了同样的问题，是跑的步数有问题吗？而且没有调整原程序 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you...