Expand description
Quality-function mu oracle — port of
IpQualityFunctionMuOracle.{hpp,cpp}. Phase 10.
The oracle picks μ_new = σ * avrg_compl by minimizing a 1-D
quality function q(σ) over σ ∈ [σ_lo, σ_up] via golden section.
The full vector-valued evaluator (which builds the trial slack /
multiplier vectors at a candidate σ and reduces them to a scalar
norm) is split into two pieces:
evaluate_quality_function— a pure-scalar reducer that takes already-computed‖·‖aggregates and combines them per the(norm, centrality, balancing)triple perIpQualityFunctionMuOracle.cpp:566-658. The vector→aggregate step is the caller’s responsibility.pick_sigma— orchestrator that mirrorsIpQualityFunctionMuOracle.cpp::CalculateMulines 329-385: picks the σ-bracket, evaluatesq(1)andq(1−ε)to decide whether to search above or below 1, then drivesgolden_section.
Wiring pick_sigma to a fully populated q(σ) evaluator —
including the centering predictor solve — is the remaining scope.
Structs§
- Quality
Function Aggregates - Per-norm aggregates feeding
evaluate_quality_function. - Quality
Function MuOracle
Enums§
Functions§
- evaluate_
quality_ function - Pure-scalar reducer corresponding to
IpQualityFunctionMuOracle.cpp::CalculateQualityFunctionlines 566-658 minus the vector→aggregate reduction. Combines the caller-provided norm aggregates per the configured(norm, centrality, balancing)triple. - golden_
section - Pure-scalar golden-section minimizer used by
QualityFunctionMuOracle::PerformGoldenSection(IpQualityFunctionMuOracle.cpp:668-790). - pick_
sigma - Sigma-bracket selection + golden-section orchestrator. Mirrors
IpQualityFunctionMuOracle.cpp::CalculateMulines 329-385.