accelerators/lbfgs.hpp File Reference

#include <alpaqa/config/config.hpp>
#include <alpaqa/export.hpp>
#include <cmath>
#include <limits>
#include <string>
#include <utility>
#include <vector>

Include dependency graph for accelerators/lbfgs.hpp:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Classes
struct	CBFGSParams< Conf >
	Cautious BFGS update. More...
struct	LBFGSParams< Conf >
	Parameters for the LBFGS class. More...
struct	LBFGSStorage< Conf >
	Layout: More...
class	LBFGS< Conf >
	Limited memory Broyden–Fletcher–Goldfarb–Shanno (L-BFGS) algorithm. More...

Namespaces
namespace	alpaqa

Enumerations
enum class	LBFGSStepSize { BasedOnExternalStepSize = 0 , BasedOnCurvature = 1 , BasedOnGradientStepSize }
	Which method to use to select the L-BFGS step size. More...

---

Class Documentation

alpaqa::LBFGSParams

struct alpaqa::LBFGSParams

Collaboration diagram for LBFGSParams< Conf >:

Class Members
length_t	memory = 10	Length of the history to keep.
real_t	min_div_fac = std::numeric_limits<real_t>::epsilon()	Reject update if \( y^\top s \le \text{min_div_fac} \cdot s^\top s \).
real_t	min_abs_s	Reject update if \( s^\top s \le \text{min_abs_s} \).
CBFGSParams< config_t >	cbfgs = {}	Parameters in the cautious BFGS update condition. \[ \frac{y^\top s}{s^\top s} \ge \epsilon \| g \|^\alpha \] See also https://epubs.siam.org/doi/10.1137/S1052623499354242
bool	force_pos_def = true	If set to true, the inverse Hessian estimate should remain definite, i.e. a check is performed that rejects the update if \( y^\top s \le \text{min_div_fac} \cdot s^\top s \). If set to false, just try to prevent a singular Hessian by rejecting the update if \( \left| y^\top s \right| \le \text{min_div_fac} \cdot s^\top s \).
LBFGSStepSize	stepsize = LBFGSStepSize::BasedOnCurvature	See also LBFGSStepSize