TMB Documentation
v1.9.11
|
Newton configuration parameters. More...
#include <newton.hpp>
Public Attributes | |
double | grad_tol |
Convergence tolerance of max gradient component. | |
bool | lowrank |
Detect an additional low rank contribution in sparse case? | |
int | max_reject |
Max number of allowed rejections. | |
int | maxit |
Max number of iterations. | |
double | mgcmax |
Consider initial guess as invalid if the max gradient component is larger than this number. | |
int | ok_exit_if_pdhess |
max_reject exceeded is convergence success provided that Hessian is PD? | |
bool | on_failure_give_warning |
Behaviour on convergence failure: Throw warning ? | |
bool | on_failure_return_nan |
Behaviour on convergence failure: Report nan-solution ? | |
double | power |
Internal parameter controlling ustep updates. | |
double | signif_abs_reduction |
Consider this absolute reduction 'significant'. | |
double | signif_rel_reduction |
Consider this relative reduction 'significant'. | |
bool | simplify |
Detect and apply 'dead gradients' simplification. | |
bool | SPA |
Modify Laplace approximation to return saddlepoint approximation? More... | |
bool | sparse |
Use sparse as opposed to dense hessian ? More... | |
double | step_tol |
Convergence tolerance of consequtive function evaluations (not yet used) | |
double | tol10 |
Convergence tolerance of consequtive function evaluations (not yet used) | |
int | trace |
Print trace info? | |
double | u0 |
Internal parameter controlling ustep updates. | |
double | ustep |
Initial step size between 0 and 1. | |
Newton configuration parameters.
Definition at line 637 of file newton.hpp.
bool newton::newton_config::SPA |
Modify Laplace approximation to return saddlepoint approximation?
For this to work the inner objective function must return the negative log MGF rather than density
Definition at line 687 of file newton.hpp.
bool newton::newton_config::sparse |
Use sparse as opposed to dense hessian ?
Using sparse=true
for problem that is actually dense have been observed to results in a slowdown factor of approximately 3. In addition, the dense factorization can be accelerated using the EIGEN_USE_MKL_ALL
preprocessor flag. On the other hand, using sparse=false
(dense) for a problem that is actually sparse can result in much bigger slowdowns.
Definition at line 669 of file newton.hpp.