OptionalfeasibilityThis same linearisation at this same penalty without the objective,
the model of the steering's step of the violation alone (their
(5.10)), for a producer whose rows do not depend on the objective.
Absent, the steering asks IProblem.convexify for it.
The model of the violation alone, with its own cold start.
OptionalrepriceReprice this same fixed linearization for steering, without rebuilding its geometry or matrices. Absent, convexify rebuilds the model.
The new exact l1 penalty, at the same iterate/objective and row domain.
A fresh model shell with the original cold/keyed-dual initialization; fixed numeric preparation may be shared sequentially.
The model's step within a region.
The region's half-width, in the caller's units.
The step.
OptionaltakenTold that the last step solved was taken, so the caller's next model can start from its solution (its multipliers, for one).
A convex model of the merit about an iterate, at one penalty.