alpaqa 1.0.0a9
Nonconvex constrained optimization
Loading...
Searching...
No Matches
Public Types | Public Member Functions | Static Public Member Functions | Public Attributes | Private Attributes | List of all members
DLProblem Class Reference

#include <alpaqa/dl/dl-problem.hpp>

Detailed Description

Class that loads a problem using dlopen.

The shared library should export a C function with the name <symbol_prefix>_register that accepts a void pointer with user data, and returns a struct of type alpaqa_problem_register_t that contains all data to represent the problem, as well as function pointers for all required operations. See C++/DLProblem/main.cpp and problems/sparse-logistic-regression.cpp for examples.

Note
Copies are shallow, they all share the same problem instance, take that into account when using multiple threads.
See also
TypeErasedProblem
alpaqa_problem_functions_t
alpaqa_problem_register_t
Examples
C++/DLProblem/main.cpp.

Definition at line 92 of file dl-problem.hpp.

+ Inheritance diagram for DLProblem:
+ Collaboration diagram for DLProblem:

Public Types

using instance_t = ExtraFuncs::instance_t
 
using Box = alpaqa::Box< config_t >
 

Public Member Functions

 DLProblem (const std::string &so_filename, std::string symbol_prefix="alpaqa_problem", void *user_param=nullptr)
 Load a problem from a shared library.
 
real_t eval_prox_grad_step (real_t γ, crvec x, crvec grad_ψ, rvec x̂, rvec p) const
 
real_t eval_f (crvec x) const
 
void eval_grad_f (crvec x, rvec grad_fx) const
 
void eval_g (crvec x, rvec gx) const
 
void eval_grad_g_prod (crvec x, crvec y, rvec grad_gxy) const
 
void eval_jac_g (crvec x, rindexvec inner_idx, rindexvec outer_ptr, rvec J_values) const
 
length_t get_jac_g_num_nonzeros () const
 
void eval_grad_gi (crvec x, index_t i, rvec grad_gi) const
 
void eval_hess_L_prod (crvec x, crvec y, real_t scale, crvec v, rvec Hv) const
 
void eval_hess_L (crvec x, crvec y, real_t scale, rindexvec inner_idx, rindexvec outer_ptr, rvec H_values) const
 
length_t get_hess_L_num_nonzeros () const
 
void eval_hess_ψ_prod (crvec x, crvec y, crvec Σ, real_t scale, crvec v, rvec Hv) const
 
void eval_hess_ψ (crvec x, crvec y, crvec Σ, real_t scale, rindexvec inner_idx, rindexvec outer_ptr, rvec H_values) const
 
length_t get_hess_ψ_num_nonzeros () const
 
real_t eval_f_grad_f (crvec x, rvec grad_fx) const
 
real_t eval_f_g (crvec x, rvec g) const
 
void eval_grad_f_grad_g_prod (crvec x, crvec y, rvec grad_f, rvec grad_gxy) const
 
void eval_grad_L (crvec x, crvec y, rvec grad_L, rvec work_n) const
 
real_t eval_ψ (crvec x, crvec y, crvec Σ, rvec ŷ) const
 
void eval_grad_ψ (crvec x, crvec y, crvec Σ, rvec grad_ψ, rvec work_n, rvec work_m) const
 
real_t eval_ψ_grad_ψ (crvec x, crvec y, crvec Σ, rvec grad_ψ, rvec work_n, rvec work_m) const
 
bool provides_eval_f () const
 
bool provides_eval_grad_f () const
 
bool provides_eval_g () const
 
bool provides_eval_grad_g_prod () const
 
bool provides_eval_jac_g () const
 
bool provides_get_jac_g_num_nonzeros () const
 
bool provides_eval_grad_gi () const
 
bool provides_eval_hess_L_prod () const
 
bool provides_eval_hess_L () const
 
bool provides_get_hess_L_num_nonzeros () const
 
bool provides_eval_hess_ψ_prod () const
 
bool provides_eval_hess_ψ () const
 
bool provides_get_hess_ψ_num_nonzeros () const
 
bool provides_eval_f_grad_f () const
 
bool provides_eval_f_g () const
 
bool provides_eval_grad_f_grad_g_prod () const
 
bool provides_eval_grad_L () const
 
bool provides_eval_ψ () const
 
bool provides_eval_grad_ψ () const
 
bool provides_eval_ψ_grad_ψ () const
 
bool provides_get_box_C () const
 
template<class Signature , class... Args>
decltype(auto) call_extra_func (const std::string &name, Args &&...args) const
 
template<class Signature , class... Args>
decltype(auto) call_extra_func (const std::string &name, Args &&...args)
 
void resize (length_t n, length_t m)
 Change the dimensions of the problem (number of decision variables and number of constaints).
 
length_t get_n () const
 Number of decision variables, n.
 
length_t get_m () const
 Number of constraints, m.
 
void eval_proj_diff_g (crvec z, rvec p) const
 
void eval_proj_multipliers (rvec y, real_t M) const
 
const Boxget_box_C () const
 
const Boxget_box_D () const
 
index_t eval_inactive_indices_res_lna (real_t γ, crvec x, crvec grad_ψ, rindexvec J) const
 
void check () const
 

Static Public Member Functions

static real_t eval_proj_grad_step_box (const Box &C, real_t γ, crvec x, crvec grad_ψ, rvec x̂, rvec p)
 Projected gradient step for rectangular box C.
 
static void eval_prox_grad_step_box_l1_impl (const Box &C, const auto &λ, real_t γ, crvec x, crvec grad_ψ, rvec x̂, rvec p)
 Proximal gradient step for rectangular box C with ℓ₁-regularization.
 
static real_t eval_prox_grad_step_box_l1 (const Box &C, const auto &λ, real_t γ, crvec x, crvec grad_ψ, rvec x̂, rvec p)
 Proximal gradient step for rectangular box C with ℓ₁-regularization.
 
static real_t eval_prox_grad_step_box_l1_scal (const Box &C, real_t λ, real_t γ, crvec x, crvec grad_ψ, rvec x̂, rvec p)
 Proximal gradient step for rectangular box C with ℓ₁-regularization.
 
static void eval_proj_multipliers_box (const Box &D, rvec y, real_t M, index_t penalty_alm_split)
 

Public Attributes

length_t n
 Number of decision variables, dimension of x.
 
length_t m
 Number of constraints, dimension of g(x) and z.
 
Box C
 Constraints of the decision variables, \( x \in C \).
 
Box D
 Other constraints, \( g(x) \in D \).
 
vec l1_reg
 \( \ell_1 \) (1-norm) regularization parameter.
 
index_t penalty_alm_split
 Components of the constraint function with indices below this number are handled using a quadratic penalty method rather than using an augmented Lagrangian method.
 

Private Attributes

std::shared_ptr< void > handle
 Handle to the shared module defining the problem.
 
std::shared_ptr< void > instance
 Problem instance created by the registration function, including the deleter to destroy it.
 
problem_functions_tfunctions = nullptr
 Pointer to the struct of function pointers for evaluating the objective, constraints, their gradients, etc.
 
ExtraFuncs extra_funcs
 Dictionary of extra functions that were registered by the problem.
 

Member Typedef Documentation

◆ instance_t

using instance_t = ExtraFuncs::instance_t

Definition at line 164 of file dl-problem.hpp.

◆ Box

using Box = alpaqa::Box<config_t>
inherited

Definition at line 19 of file box-constr-problem.hpp.

Constructor & Destructor Documentation

◆ DLProblem()

DLProblem ( const std::string &  so_filename,
std::string  symbol_prefix = "alpaqa_problem",
void *  user_param = nullptr 
)

Load a problem from a shared library.

Parameters
so_filenameFilename of the shared library to load.
symbol_prefixPrefix of the registration function.
user_paramPointer to custom user data to pass to the registration function.

Definition at line 70 of file dl-problem.cpp.

Member Function Documentation

◆ eval_prox_grad_step()

auto eval_prox_grad_step ( real_t  γ,
crvec  x,
crvec  grad_ψ,
rvec  ,
rvec  p 
) const

Definition at line 124 of file dl-problem.cpp.

+ Here is the call graph for this function:

◆ eval_f()

auto eval_f ( crvec  x) const

Definition at line 133 of file dl-problem.cpp.

◆ eval_grad_f()

auto eval_grad_f ( crvec  x,
rvec  grad_fx 
) const

Definition at line 134 of file dl-problem.cpp.

◆ eval_g()

auto eval_g ( crvec  x,
rvec  gx 
) const

Definition at line 135 of file dl-problem.cpp.

◆ eval_grad_g_prod()

auto eval_grad_g_prod ( crvec  x,
crvec  y,
rvec  grad_gxy 
) const

Definition at line 136 of file dl-problem.cpp.

◆ eval_jac_g()

auto eval_jac_g ( crvec  x,
rindexvec  inner_idx,
rindexvec  outer_ptr,
rvec  J_values 
) const

Definition at line 138 of file dl-problem.cpp.

◆ get_jac_g_num_nonzeros()

auto get_jac_g_num_nonzeros ( ) const

Definition at line 139 of file dl-problem.cpp.

◆ eval_grad_gi()

auto eval_grad_gi ( crvec  x,
index_t  i,
rvec  grad_gi 
) const

Definition at line 137 of file dl-problem.cpp.

◆ eval_hess_L_prod()

auto eval_hess_L_prod ( crvec  x,
crvec  y,
real_t  scale,
crvec  v,
rvec  Hv 
) const

Definition at line 140 of file dl-problem.cpp.

◆ eval_hess_L()

auto eval_hess_L ( crvec  x,
crvec  y,
real_t  scale,
rindexvec  inner_idx,
rindexvec  outer_ptr,
rvec  H_values 
) const

Definition at line 141 of file dl-problem.cpp.

◆ get_hess_L_num_nonzeros()

auto get_hess_L_num_nonzeros ( ) const

Definition at line 142 of file dl-problem.cpp.

◆ eval_hess_ψ_prod()

auto eval_hess_ψ_prod ( crvec  x,
crvec  y,
crvec  Σ,
real_t  scale,
crvec  v,
rvec  Hv 
) const

Definition at line 143 of file dl-problem.cpp.

◆ eval_hess_ψ()

auto eval_hess_ψ ( crvec  x,
crvec  y,
crvec  Σ,
real_t  scale,
rindexvec  inner_idx,
rindexvec  outer_ptr,
rvec  H_values 
) const

Definition at line 144 of file dl-problem.cpp.

◆ get_hess_ψ_num_nonzeros()

auto get_hess_ψ_num_nonzeros ( ) const

Definition at line 145 of file dl-problem.cpp.

◆ eval_f_grad_f()

auto eval_f_grad_f ( crvec  x,
rvec  grad_fx 
) const

Definition at line 146 of file dl-problem.cpp.

◆ eval_f_g()

auto eval_f_g ( crvec  x,
rvec  g 
) const

Definition at line 147 of file dl-problem.cpp.

◆ eval_grad_f_grad_g_prod()

auto eval_grad_f_grad_g_prod ( crvec  x,
crvec  y,
rvec  grad_f,
rvec  grad_gxy 
) const

Definition at line 148 of file dl-problem.cpp.

◆ eval_grad_L()

auto eval_grad_L ( crvec  x,
crvec  y,
rvec  grad_L,
rvec  work_n 
) const

Definition at line 149 of file dl-problem.cpp.

◆ eval_ψ()

auto eval_ψ ( crvec  x,
crvec  y,
crvec  Σ,
rvec  ŷ 
) const

Definition at line 150 of file dl-problem.cpp.

◆ eval_grad_ψ()

auto eval_grad_ψ ( crvec  x,
crvec  y,
crvec  Σ,
rvec  grad_ψ,
rvec  work_n,
rvec  work_m 
) const

Definition at line 151 of file dl-problem.cpp.

◆ eval_ψ_grad_ψ()

auto eval_ψ_grad_ψ ( crvec  x,
crvec  y,
crvec  Σ,
rvec  grad_ψ,
rvec  work_n,
rvec  work_m 
) const

Definition at line 152 of file dl-problem.cpp.

◆ provides_eval_f()

bool provides_eval_f ( ) const

Definition at line 154 of file dl-problem.cpp.

◆ provides_eval_grad_f()

bool provides_eval_grad_f ( ) const

Definition at line 155 of file dl-problem.cpp.

◆ provides_eval_g()

bool provides_eval_g ( ) const

Definition at line 156 of file dl-problem.cpp.

◆ provides_eval_grad_g_prod()

bool provides_eval_grad_g_prod ( ) const

Definition at line 157 of file dl-problem.cpp.

◆ provides_eval_jac_g()

bool provides_eval_jac_g ( ) const

Definition at line 158 of file dl-problem.cpp.

◆ provides_get_jac_g_num_nonzeros()

bool provides_get_jac_g_num_nonzeros ( ) const

Definition at line 159 of file dl-problem.cpp.

◆ provides_eval_grad_gi()

bool provides_eval_grad_gi ( ) const

Definition at line 160 of file dl-problem.cpp.

◆ provides_eval_hess_L_prod()

bool provides_eval_hess_L_prod ( ) const

Definition at line 161 of file dl-problem.cpp.

◆ provides_eval_hess_L()

bool provides_eval_hess_L ( ) const

Definition at line 162 of file dl-problem.cpp.

◆ provides_get_hess_L_num_nonzeros()

bool provides_get_hess_L_num_nonzeros ( ) const

Definition at line 163 of file dl-problem.cpp.

◆ provides_eval_hess_ψ_prod()

bool provides_eval_hess_ψ_prod ( ) const

Definition at line 164 of file dl-problem.cpp.

◆ provides_eval_hess_ψ()

bool provides_eval_hess_ψ ( ) const

Definition at line 165 of file dl-problem.cpp.

◆ provides_get_hess_ψ_num_nonzeros()

bool provides_get_hess_ψ_num_nonzeros ( ) const

Definition at line 166 of file dl-problem.cpp.

◆ provides_eval_f_grad_f()

bool provides_eval_f_grad_f ( ) const

Definition at line 167 of file dl-problem.cpp.

◆ provides_eval_f_g()

bool provides_eval_f_g ( ) const

Definition at line 168 of file dl-problem.cpp.

◆ provides_eval_grad_f_grad_g_prod()

bool provides_eval_grad_f_grad_g_prod ( ) const

Definition at line 169 of file dl-problem.cpp.

◆ provides_eval_grad_L()

bool provides_eval_grad_L ( ) const

Definition at line 170 of file dl-problem.cpp.

◆ provides_eval_ψ()

bool provides_eval_ψ ( ) const

Definition at line 171 of file dl-problem.cpp.

◆ provides_eval_grad_ψ()

bool provides_eval_grad_ψ ( ) const

Definition at line 172 of file dl-problem.cpp.

◆ provides_eval_ψ_grad_ψ()

bool provides_eval_ψ_grad_ψ ( ) const

Definition at line 173 of file dl-problem.cpp.

◆ provides_get_box_C()

bool provides_get_box_C ( ) const

Definition at line 174 of file dl-problem.cpp.

◆ call_extra_func() [1/2]

decltype(auto) call_extra_func ( const std::string &  name,
Args &&...  args 
) const
inline

Definition at line 167 of file dl-problem.hpp.

◆ call_extra_func() [2/2]

decltype(auto) call_extra_func ( const std::string &  name,
Args &&...  args 
)
inline

Definition at line 175 of file dl-problem.hpp.

+ Here is the call graph for this function:

◆ resize()

void resize ( length_t  n,
length_t  m 
)
inlineinherited

Change the dimensions of the problem (number of decision variables and number of constaints).

Destructive: resizes and/or resets the members C, D, l1_reg and penalty_alm_split.

Definition at line 41 of file box-constr-problem.hpp.

◆ get_n()

length_t get_n ( ) const
inlineinherited

Number of decision variables, n.

Definition at line 75 of file box-constr-problem.hpp.

◆ get_m()

length_t get_m ( ) const
inlineinherited

Number of constraints, m.

Definition at line 77 of file box-constr-problem.hpp.

◆ eval_proj_grad_step_box()

static real_t eval_proj_grad_step_box ( const Box C,
real_t  γ,
crvec  x,
crvec  grad_ψ,
rvec  ,
rvec  p 
)
inlinestaticinherited

Projected gradient step for rectangular box C.

\[ \begin{aligned} \hat x &= \Pi_C(x - \gamma\nabla\psi(x)) \\ p &= \hat x - x \\ &= \max(\underline x - x, \;\min(-\gamma\nabla\psi(x), \overline x - x) \end{aligned} \]

Definition at line 84 of file box-constr-problem.hpp.

◆ eval_prox_grad_step_box_l1_impl()

static void eval_prox_grad_step_box_l1_impl ( const Box C,
const auto &  λ,
real_t  γ,
crvec  x,
crvec  grad_ψ,
rvec  ,
rvec  p 
)
inlinestaticinherited

Proximal gradient step for rectangular box C with ℓ₁-regularization.

\[ \begin{aligned} h(x) &= \|x\|_1 + \delta_C(x) \\ \hat x &= \prox_{\gamma h}(x - \gamma\nabla\psi(x)) \\ &= -\max\big( x - \overline x, \;\min\big( x - \underline x, \;\min\big( \gamma(\nabla\psi(x) + \lambda), \;\max\big( \gamma(\nabla\psi(x) - \lambda), x \big) \big) \big) \big) \end{aligned} \]

Definition at line 107 of file box-constr-problem.hpp.

◆ eval_prox_grad_step_box_l1()

static real_t eval_prox_grad_step_box_l1 ( const Box C,
const auto &  λ,
real_t  γ,
crvec  x,
crvec  grad_ψ,
rvec  ,
rvec  p 
)
inlinestaticinherited

Proximal gradient step for rectangular box C with ℓ₁-regularization.

\[ \begin{aligned} h(x) &= \|x\|_1 + \delta_C(x) \\ \hat x &= \prox_{\gamma h}(x - \gamma\nabla\psi(x)) \\ &= -\max\big( x - \overline x, \;\min\big( x - \underline x, \;\min\big( \gamma(\nabla\psi(x) + \lambda), \;\max\big( \gamma(\nabla\psi(x) - \lambda), x \big) \big) \big) \big) \end{aligned} \]

Definition at line 116 of file box-constr-problem.hpp.

◆ eval_prox_grad_step_box_l1_scal()

static real_t eval_prox_grad_step_box_l1_scal ( const Box C,
real_t  λ,
real_t  γ,
crvec  x,
crvec  grad_ψ,
rvec  ,
rvec  p 
)
inlinestaticinherited

Proximal gradient step for rectangular box C with ℓ₁-regularization.

\[ \begin{aligned} h(x) &= \|x\|_1 + \delta_C(x) \\ \hat x &= \prox_{\gamma h}(x - \gamma\nabla\psi(x)) \\ &= -\max\big( x - \overline x, \;\min\big( x - \underline x, \;\min\big( \gamma(\nabla\psi(x) + \lambda), \;\max\big( \gamma(\nabla\psi(x) - \lambda), x \big) \big) \big) \big) \end{aligned} \]

Definition at line 123 of file box-constr-problem.hpp.

◆ eval_proj_diff_g()

void eval_proj_diff_g ( crvec  z,
rvec  p 
) const
inlineinherited
See also
TypeErasedProblem::eval_proj_diff_g

Definition at line 142 of file box-constr-problem.hpp.

◆ eval_proj_multipliers_box()

static void eval_proj_multipliers_box ( const Box D,
rvec  y,
real_t  M,
index_t  penalty_alm_split 
)
inlinestaticinherited

Definition at line 144 of file box-constr-problem.hpp.

◆ eval_proj_multipliers()

void eval_proj_multipliers ( rvec  y,
real_t  M 
) const
inlineinherited

◆ get_box_C()

const Box & get_box_C ( ) const
inlineinherited
See also
TypeErasedProblem::get_box_C

Definition at line 171 of file box-constr-problem.hpp.

◆ get_box_D()

const Box & get_box_D ( ) const
inlineinherited
See also
TypeErasedProblem::get_box_D

Definition at line 173 of file box-constr-problem.hpp.

◆ eval_inactive_indices_res_lna()

index_t eval_inactive_indices_res_lna ( real_t  γ,
crvec  x,
crvec  grad_ψ,
rindexvec  J 
) const
inlineinherited

◆ check()

void check ( ) const
inlineinherited
See also
TypeErasedProblem::check

Definition at line 213 of file box-constr-problem.hpp.

Member Data Documentation

◆ handle

std::shared_ptr<void> handle
private

Handle to the shared module defining the problem.

Definition at line 107 of file dl-problem.hpp.

◆ instance

std::shared_ptr<void> instance
private

Problem instance created by the registration function, including the deleter to destroy it.

Definition at line 110 of file dl-problem.hpp.

◆ functions

problem_functions_t* functions = nullptr
private

Pointer to the struct of function pointers for evaluating the objective, constraints, their gradients, etc.

Definition at line 113 of file dl-problem.hpp.

◆ extra_funcs

ExtraFuncs extra_funcs
private

Dictionary of extra functions that were registered by the problem.

Definition at line 115 of file dl-problem.hpp.

◆ n

length_t n
inherited

Number of decision variables, dimension of x.

Definition at line 22 of file box-constr-problem.hpp.

◆ m

length_t m
inherited

Number of constraints, dimension of g(x) and z.

Definition at line 24 of file box-constr-problem.hpp.

◆ C

Box C
inherited

Constraints of the decision variables, \( x \in C \).

Definition at line 59 of file box-constr-problem.hpp.

◆ D

Box D
inherited

Other constraints, \( g(x) \in D \).

Definition at line 61 of file box-constr-problem.hpp.

◆ l1_reg

vec l1_reg
inherited

\( \ell_1 \) (1-norm) regularization parameter.

Possible dimensions are: \( 0 \) (no regularization), \( 1 \) (a single scalar factor), or \( n \) (a different factor for each variable).

Definition at line 66 of file box-constr-problem.hpp.

◆ penalty_alm_split

index_t penalty_alm_split
inherited

Components of the constraint function with indices below this number are handled using a quadratic penalty method rather than using an augmented Lagrangian method.

Specifically, the Lagrange multipliers for these components (which determine the shifts in ALM) are kept at zero.

Definition at line 72 of file box-constr-problem.hpp.


The documentation for this class was generated from the following files: