hard boundary conditions and anchors

[pkpilly]

It is not clear how data (via icbc.PointSetBC and anchors) can be used for training the PINN when hard boundary conditions are enforced (via apply_output_transforms). Any guidance will be appreciated.

[praksharma]

Sampling and output transform are two separate things. Why is that a problem? Put your anchor points (manual PDE points) and BC points (manual BC points) as

data = dde.data.TimePDE(
    geomtime,
    pde,
    [bc, ic, observe_y],
    num_domain=40,
    num_boundary=20,
    num_initial=10,
    anchors=observe_x,
    solution=func,
    num_test=10000,
)

Hard constraining is just a function binded to the output neurons. Just use them from the tutorials.

net.apply_output_transform(FUNCTION)

Check this. They are independent.