DCT Lexicon · Robustness
Robust optimization (max-min plan)
A robust problem declares a bounded set of outside events and chooses the decision that maximizes its worst value over the set, keeping every commitment for every event in it.
Vol. II, Definition 10.1, in the book's words
In depth
In the book
A robust problem declares a bounded set of outside events and chooses a decision to maximize its worst value over , keeping every commitment for every event in :
Vol. II, Definition 10.1 (Robust Problem and Graded Declaration), verbatim
Scope and limits
Robust decisions answer to the set, not to the odds: a set too small gives a guarantee that the future can step around, and a set too large buys protection against events nobody fears.
Vol. II, Section 10.1, Scope and Limits, verbatim
Sets and odds
If a law gives the declared set probability at least , a robust decision keeps the commitment with probability at least . … Part (iii) is the bridge to Chapter 9: a set that holds 95 percent of the law’s futures promises a 5 percent appetite, conservatively, since a robust decision may survive many futures outside its set.
Vol. II, Theorem 10.1(iii) and its Enterprise Interpretation, verbatim
Common misreading
A common mistake is to read a robust guarantee as a guarantee about the future rather than about the set.
Vol. II, LOS 10.1, verbatim
How AXIOM uses it
AXIOM solves the max-min plan on the 95% box derived from the declared law, with HiGHS, and shows it beside Optimize's plan: navy bars are the signed plan, green bars the robust plan. It certifies the worst case on the set, and is statistical under the law.
Drafted for the author's approval, from AXIOM's method passport "Robust plan and the cost of doubt" (D-063)
In the book: Vol. II, Ch. 10: Neuro-Fuzzy Robust Enterprise Optimization
Related: Price of robustness (cost of doubt) · Robust Viability · Wasserstein Distributionally Robust Optimization · Kullback–Leibler divergence
V2.Def.10.1 · V2.Thm.10.1


