You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
First, congratulations on the excellent review on the exciting topic! I would like to introduce my paper as a geoscience example for discrete adjoints and reverse AD.
In our latest research, we developed a fully-differentiable solver for a nonlinear time-dependent PDE on JAX, and estimate the constitutive relations via neural networks embedded in the PDE. Rather than implementing discrete adjoint methods, we implemented implicit differentiation as a custom derivative rule for the nonlinear solver (probably equivalent to Jilia's way mentioned at the end of section 3.9.2?).
Aside from the paper, in Chapter 4 of my PhD thesis, I compared discrete and continuous adjoint methods for the same PDE.
Hope they are helpful!
The text was updated successfully, but these errors were encountered:
Hi,
First, congratulations on the excellent review on the exciting topic! I would like to introduce my paper as a geoscience example for discrete adjoints and reverse AD.
In our latest research, we developed a fully-differentiable solver for a nonlinear time-dependent PDE on JAX, and estimate the constitutive relations via neural networks embedded in the PDE. Rather than implementing discrete adjoint methods, we implemented implicit differentiation as a custom derivative rule for the nonlinear solver (probably equivalent to Jilia's way mentioned at the end of section 3.9.2?).
Aside from the paper, in Chapter 4 of my PhD thesis, I compared discrete and continuous adjoint methods for the same PDE.
Hope they are helpful!
The text was updated successfully, but these errors were encountered: