mir-optim ~master (2021-05-13T12:53:52Z)

LeastSquaresSettings

mir optim least_squares

Least-Squares iteration settings.

struct LeastSquaresSettings (
T
) if (
is(T == double) ||
is(T == float)
) {
import mir.optim.boxcqp;;
import mir.math.common : sqrt;;
import mir.math.constant : GoldenRatio;;
import lapack : lapackint;;
uint maxIterations;
uint maxAge;
T jacobianEpsilon;
T absTolerance;
T relTolerance;
T gradTolerance;
T maxGoodResidual;
T maxStep;
T maxLambda;
T minLambda;
T minStepQuality;
T goodStepQuality;
T lambdaIncrease;
T lambdaDecrease;
BoxQPSettings!T qpSettings;
}

Members

Variables

absTolerance
T absTolerance;

Absolute tolerance for step size, L2 norm

goodStepQuality
T goodStepQuality;

for steps above this quality, the trust region is expanded

gradTolerance
T gradTolerance;

Absolute tolerance for gradient, L-inf norm

jacobianEpsilon
T jacobianEpsilon;

epsilon for finite difference Jacobian approximation

lambdaDecrease
T lambdaDecrease;

lambda is multiplied by this factor after good quality steps

lambdaIncrease
T lambdaIncrease;

lambda is multiplied by this factor after step below min quality

maxAge
uint maxAge;

Maximum jacobian model age (0 for default selection)

maxGoodResidual
T maxGoodResidual;

The algorithm stops iteration when the residual value is less or equal to maxGoodResidual.

maxIterations
uint maxIterations;

Maximum number of iterations

maxLambda
T maxLambda;

minimum trust region radius

maxStep
T maxStep;

maximum norm of iteration step

minLambda
T minLambda;

maximum trust region radius

minStepQuality
T minStepQuality;

for steps below this quality, the trust region is shrinked

qpSettings
BoxQPSettings!T qpSettings;

Bound constrained convex quadratic problem settings

relTolerance
T relTolerance;

Relative tolerance for step size, L2 norm

Meta

Source

See Implementation

mir optim least_squares aliasesenumsfunctionsstructs

Page generated by adrdox