line::mdd::MddMcdOptions Struct Reference

Knobs of the level iteration in mdd_mcd. More...

#include <line/api/mdd/mdd_types.h>

Collaboration diagram for line::mdd::MddMcdOptions:

Public Attributes
double	tol = 1e-12
	Convergence tolerance on the level marginals.
int	maxiter = 500
	Maximum coupled sweeps before the iteration is declared non-convergent.
std::vector< std::vector< double > >	initpik
	The reference's 'verbose' knob is NOT carried: it is a console trace of the level sizes and the iteration count, and this port keeps the api layer silent, as infer_lqn_ekf does.

Detailed Description

Knobs of the level iteration in mdd_mcd.

The defaults are deliberately much tighter than a solver-level fixed-point tolerance: the level iteration is an INNER numerical solve and mdd_mcd verifies the population invariant at 1e-6, so a loose tolerance converges short of the fixed point and trips that guard. Do not wire an AMVA-sized iter_tol into these.

Definition at line 212 of file mdd_types.h.

Member Data Documentation

initpik

std::vector<std::vector<double> > line::mdd::MddMcdOptions::initpik

The reference's 'verbose' knob is NOT carried: it is a console trace of the level sizes and the iteration count, and this port keeps the api layer silent, as infer_lqn_ekf does.

MddMcdResult returns level_sizes and iters, so the same numbers are available to a caller that wants them. Optional warm-start level vectors, one per paper level; empty to start uniform.

Definition at line 224 of file mdd_types.h.

maxiter

int line::mdd::MddMcdOptions::maxiter = 500

Maximum coupled sweeps before the iteration is declared non-convergent.

Definition at line 216 of file mdd_types.h.

tol

double line::mdd::MddMcdOptions::tol = 1e-12

Convergence tolerance on the level marginals.

Definition at line 214 of file mdd_types.h.

---

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

• include/line/api/mdd/mdd_types.h