-
**Please describe the problem you are trying to solve.**
Quasi newton acceleration is only applied to the last timestep of the waveform iteration currently. This yields slow convergence for higher or…
-
Is there any interest in adding a quasi-Newton based optimizer? I was thinking of porting over:
https://github.com/tensorflow/probability/blob/master/tensorflow_probability/python/optimizer/bfgs.py…
-
Currently, the elliptic grid solver defaults to applying a successive line over relaxation for several iterations, then applying a full newton solve to converge the grid down to 1e-14.
The jacobian…
-
Generate flamegraphs for the quasi-Newton and relaxation approaches and assess if optimisations can be made.
-
NewtonImplicitSolver is now recalculating the Jacobian and factorizing it in every iteration of the Newton algorithm. Using a Quasi-Newton approach (and e.g. only update the Jacobian from time to time…
-
In the Quasi-Newton methods, especially close to the optimum, it often gets to the point that the search direction is very nearly perpendicular to the gradient. This harms our convergence by at least …
-
We have some acceleration schemes that are no quasi-Newton schemes. However, they still write the QN-related columns `QNColumns`, `DeletedQNColumns`, and `DroppedQNColumns` in `*-iterations.log`. It w…
-
I am trying to benchmark solvers on a Diagonal Quasi-Newton Model. I get two issues:
1. I get `type DiagonalQNModel has no field counters`
2. When giving a Quasi-Newton Model to Ipopt, I get `Meth…
-
I think the try/catch block [in the structure's quasi-static solver](https://github.com/exadg/exadg/blob/b9bc9fec2bd1f432c6ea6da64181e3d2002d64bd/include/exadg/structure/time_integration/driver_quasi_…
-
We can get some nice speedup if we use `LinearAlgebra.BLAS.ger!` to do the rank-1 updates of the approximate Hessian or inverse Hessian in quasi-Newton methods instead of `H = H + (w * w') / θ` which …