Closed HenKlei closed 4 years ago
@sdrave Maybe you could have a look at 037b576? I made changes addressing this issue as well as #4.
This is basically right, but the result of solve
always should be a NumpyVectorArray
not a plain numpy array. You should also define solution_space
as NumpyVectorSpace(len(reduced_basis))
. Then you can write self.solution_space.make_array(u)
.
Since I just removed reduced_basis
from NeuralNetworkModel
, I will use the output dimension of the neural network instead of len(reduced_basis)
.
Should be fixed in 25d4f8a.
LGTM.
The model should return the coordinates in the reduced basis and not their reconstruction in the high dimensional state.