forxltk

@happyzhouch Actually, the result is good. But you reshape `y_pred` in wrong way. Please use the following code. ``` y_pred_2d= y_pred.reshape(xx.shape) plt.figure() plt.xlabel("x") plt.ylabel("y") plt.pcolor(xx, yy, y_pred_2d, cmap=plt.cm.jet) plt.colorbar() plt.savefig('./y_pred.png')...

@happyzhouch I am not an expert in heat problem. I don't know how the result is strange. But your code is quite strange. By default, `i` and `j` are `0`...

``` def init_func(X): x = X[:, 0:1] y = X[:, 1:2] out = np.zeros((len(X),1)) for count, y_ in enumerate(y): if (y_==0.4): out[count] = 1 return out ``` Use `np.isclose()`, not...

Maybe you should balance the losses, and you can use hard BC here.

The `metrics` in `model.compile` requires a reference solution in `dde.data.PDE`.

> Ah, I see - thank you for the comment @forxltk ! I am slightly concerned about this, is this not providing the answer we are hoping to determine via...

Maybe you can read the DeepONet [https://deepxde.readthedocs.io/en/latest/user/research.html#deeponet-mionet-deepm-mnet](url)

I think it is doable. But I have found some bug in your code. 1. The definition of `4th-order derivatives` should be something like `dy_xxxx=dde.grad.hessian(dy_xx, x, i=0, j=0)`. 2. You...

`geom` should be `geomtime` in the BC.