AugmentedLagrangianSolverOptions

Options & Configuration · AiDotNet.Models.Options

Configuration for the augmented Lagrangian solver.

Inheritance: ModelOptions → AugmentedLagrangianSolverOptions

For Beginners

These settings control how accurately the solver satisfies constraints and how much work it may spend doing so. The defaults are suitable for most smooth constrained problems.

How It Works

Reference: The defaults follow the method of multipliers as presented in J. Nocedal and S. J. Wright, Numerical Optimization (2nd ed., Springer 2006), Framework 17.3 and Algorithm 17.4.

Constructors

Constructor Summary
AugmentedLagrangianSolverOptions() Initializes the options with documented defaults.
AugmentedLagrangianSolverOptions(AugmentedLagrangianSolverOptions other) Initializes the options by copying another configuration.

Properties

Property Summary
double FeasibilityTolerance { get; set; } Gets or sets the constraint-violation tolerance that counts as feasible.
double InitialPenalty { get; set; } Gets or sets the initial penalty weight on constraint violation.
double MaximumPenalty { get; set; } Gets or sets the ceiling on the penalty weight.
double PenaltyGrowthFactor { get; set; } Gets or sets the factor by which the penalty grows when an outer iteration fails to reduce the constraint violation enough.
double RequiredViolationReduction { get; set; } Gets or sets the fraction by which the constraint violation must fall for an outer iteration to count as progress.
double StationarityTolerance { get; set; } Gets or sets the tolerance on the augmented Lagrangian’s gradient at the inner solution.
int MaxInnerIterations { get; set; } Gets or sets the maximum number of iterations the inner unconstrained solver may take per outer iteration.
int MaxOuterIterations { get; set; } Gets or sets the maximum number of outer iterations.