Originally posted by **bottle88** May 11, 2024
Dear All, I have a question of using the implementation of PFNN.
I have a 2D heat equation where T(x,y,t) is the solution. And I have a inverse **varaible** parameter in my BC. The parameter only applies on a particular boundary. so it is supposed to dependent on x only.
In the PFNN implementation. The First NN has x,y,t as inputs and T(x,y,t) as ouput , and second NN also has x,y,t as input and C(x,y,t) as output, as far I understand. But I need the second NN to have x as input and C(x) as output. How do I do this
I have seen the poison equation example, where they had unknown source field as output, and tried to implement it. But the function prediction was not at all correct. https://deepxde.readthedocs.io/en/stable/demos/pinn_inverse/elliptic.inverse.field.html.
This is how I had defined, seeing the example given.
`layer_size = [3, [30,30], [30,30], [30,30], [30,30], [30,30], 2]
activation = "tanh"
initializer = "Glorot uniform"
net = dde.nn.PFNN(layer_size, activation, initializer)`.
I am having doubt on what do I need to change to implement according to my condition. I am confused.
I am eager to hear your response. Thankyou in Advance.
Discussed in https://github.com/lululxvi/deepxde/discussions/1733