|
template<class Base , class Vector_set >
void RevHesSweep |
( |
size_t |
n, |
|
|
size_t |
numvar, |
|
|
player< Base > * |
play, |
|
|
Vector_set & |
for_jac_sparse, |
|
|
bool * |
RevJac, |
|
|
Vector_set & |
rev_hes_sparse |
|
) |
| |
Given the forward Jacobian sparsity pattern for all the variables, and the reverse Jacobian sparsity pattern for the dependent variables, RevHesSweep computes the Hessian sparsity pattern for all the independent variables.
- Template Parameters:
-
Base | base type for the operator; i.e., this operation sequence was recorded using AD< Base > and computations by this routine are done using type Base. |
Vector_set | is the type used for vectors of sets. It can be either sparse_pack or sparse_set . |
- Parameters:
-
n | is the number of independent variables on the tape. |
numvar | is the total number of variables on the tape; i.e., play->num_rec_var(). This is also the number of rows in the entire sparsity pattern rev_hes_sparse. |
play | The information stored in play is a recording of the operations corresponding to a function
where is the number of independent variables and is the number of dependent variables. The object play is effectly constant. It is not declared const because while playing back the tape the object play holds information about the currentl location with in the tape and this changes during playback. |
for_jac_sparse | For i = 0 , ... , numvar - 1, (for all the variables on the tape), the forward Jacobian sparsity pattern for the variable with index i corresponds to the set with index i in for_jac_sparse. |
RevJac | Input: For i = 0, ... , numvar - 1 the if the variable with index i on the tape is an dependent variable and included in the Hessian, RevJac[ i ] is equal to true, otherwise it is equal to false.
Output: The values in RevJac upon return are not specified; i.e., it is used for temporary work space. |
rev_hes_sparse | The reverse Hessian sparsity pattern for the variable with index i corresponds to the set with index i in rev_hes_sparse.
Input: For i = 0 , ... , numvar - 1 the reverse Hessian sparsity pattern for the variable with index i is empty.
Output: For j = 1 , ... , n, the reverse Hessian sparsity pattern for the independent dependent variable with index (j-1) is given by the set with index j in rev_hes_sparse. The values in the rest of rev_hes_sparse are not specified; i.e., they are used for temporary work space. |
Definition at line 102 of file rev_hes_sweep.hpp.
Referenced by RevSparseHesBool(), and RevSparseHesSet().
|