iPiano
An implementation of the iPiano algorithms for non-convex and non-smooth optimization.
 All Classes Functions Variables
Public Attributes | List of all members
iPiano::Options Struct Reference

Options of algorithm. More...

#include <ipiano.h>

Inheritance diagram for iPiano::Options:
nmiPiano::Options

Public Attributes

float beta_0m1 = 0.5f
 beta_0m1 to initialize alpha_0m1, delta_0m1 and gamma_0m1 according to Equations (21) and (22).
 
float c_1 = 1e-8
 Fixed c_1.
 
float c_2 = 1e-8
 Fixed c_2.
 
int steps = 10000
 Number of dicsrete steps for alpha_n and beta_n to try.
 
- Public Attributes inherited from nmiPiano::Options
Eigen::MatrixXf x_0
 Initial iterate.
 
unsigned int max_iter
 Maximum number of iterations.
 
float beta = 0.5f
 Fixed beta in [0, 1).
 
float eta = 1.05f
 Fixed eta for backtracking the local lipschitz constant.
 
float L_0m1 = 1.f
 Initialization of loca Lipschitz.
 
bool BOUND_L_N = false
 Whether to bound estimated Lipschitz constant below by the given L_n.
 
float epsilon = 0
 Termination criterion; stop if Delta_n smaller than epsilon.
 

Detailed Description

Options of algorithm.


The documentation for this struct was generated from the following file: