expression matrices. More...
#include <CouenneExprHess.hpp>
Public Member Functions | |
ExprHess () | |
empty constructor More... | |
ExprHess (CouenneProblem *) | |
code for refilling jacobian More... | |
ExprHess (const ExprHess &) | |
copy constructor More... | |
ExprHess & | operator= (const ExprHess &) |
code for refilling jacobian More... | |
ExprHess * | clone () |
Cloning operator. More... | |
~ExprHess () | |
Destructor. More... | |
int | nnz () |
int * | iRow () |
int * | jCol () |
int * | numL () |
int ** | lamI () |
expression *** | expr () |
Private Attributes | |
int | nnz_ |
number of (symbolic) nonzeroes More... | |
int * | iRow_ |
row indices (read this way by eval_h) More... | |
int * | jCol_ |
col indices More... | |
int * | numL_ |
There are m+1 (m constraints + 1 obj) components: More... | |
int ** | lamI_ |
vector of indices in the lambda vector whose constraint has nonzero entry in this position of the hessian More... | |
expression *** | expr_ |
list of lists of pointers to expression More... | |
expression matrices.
Used to evaluate the Hessian of the Lagrangian function at an optimal solution of the NLP
Definition at line 21 of file CouenneExprHess.hpp.
ExprHess::ExprHess | ( | ) |
empty constructor
Definition at line 22 of file CouenneExprHess.cpp.
ExprHess::ExprHess | ( | CouenneProblem * | p | ) |
code for refilling jacobian
for each j in (obj,con) create j->deplist()
for each variable i create dense row for each j in (obj,con) for k in j->deplist(): k<=i if d^2(j)/(di dj) nonzero add term to list [i,j] sparsify row
for each variable, fill a row of the hessian
fill term for objective
Definition at line 111 of file CouenneExprHess.cpp.
ExprHess::ExprHess | ( | const ExprHess & | rhs | ) |
copy constructor
Definition at line 33 of file CouenneExprHess.cpp.
ExprHess::~ExprHess | ( | ) |
Destructor.
Definition at line 74 of file CouenneExprHess.cpp.
code for refilling jacobian
Definition at line 38 of file CouenneExprHess.cpp.
ExprHess * ExprHess::clone | ( | ) |
Cloning operator.
Definition at line 69 of file CouenneExprHess.cpp.
|
inline |
Definition at line 63 of file CouenneExprHess.hpp.
|
inline |
Definition at line 64 of file CouenneExprHess.hpp.
|
inline |
Definition at line 65 of file CouenneExprHess.hpp.
|
inline |
Definition at line 66 of file CouenneExprHess.hpp.
|
inline |
Definition at line 67 of file CouenneExprHess.hpp.
|
inline |
Definition at line 69 of file CouenneExprHess.hpp.
|
private |
number of (symbolic) nonzeroes
Definition at line 25 of file CouenneExprHess.hpp.
|
private |
row indices (read this way by eval_h)
Definition at line 26 of file CouenneExprHess.hpp.
|
private |
col indices
Definition at line 27 of file CouenneExprHess.hpp.
|
private |
There are m+1 (m constraints + 1 obj) components:
Implementing a FP requires adding one for gg', the gradient again being set up at the beginning (at least its expression members are known).
This can simply be hacked by the FP itself. Same for the changed hessian, simply replace the CouenneProblem's objective with sum (objective, norm)
Actually, we could do the gg' trick by replacing the objective with sum (objective, norm, gg') size of each lamI_
Definition at line 44 of file CouenneExprHess.hpp.
|
private |
vector of indices in the lambda vector whose constraint has nonzero entry in this position of the hessian
Definition at line 46 of file CouenneExprHess.hpp.
|
private |
list of lists of pointers to expression
Definition at line 50 of file CouenneExprHess.hpp.