Linear Optimization

Motivation

Chapter 8 established an exact correspondence between lift-and-project cuts and simple disjunctive cuts, but its practical payoff is what this chapter develops: the cut-generating LP (CGLP)_k never needs to be formulated or solved on its own. Every pivot of (CGLP)_k can instead be mimicked directly on the much smaller simplex tableau of the original LP relaxation — replacing a large auxiliary linear program with bookkeeping on a tableau the solver already has. This chapter works out that correspondence at the level of individual pivots: which tableau pivot improves the resulting cut, and by how much, answered entirely in terms of ordinary tableau coefficients and two closed-form evaluation functions.

Setting

S:={1,…,m+p} and N:={m+p+1,…,m+p+n} index the surplus and structural variables of (LP) respectively — giving a direct correspondence between (LP)'s own variables and the surplus variables of Ãx≥b̃. For a basic solution with nonbasic set J (row set M1∪M2 from Chapter 8), Â:=Ã_J is the resulting nonsingular submatrix, and row k of the tableau reads x_k+ Σ_{j∈J}ā_{kj}s_j=ā_{k0}. Adding γ times row i to row k gives the composite row (9.10), x_k+γx_i+Σ_{j∈J}(ā_{kj}+γā_{ij})s_j=ā_{k0}+γā_{i0}, from which a new simple disjunctive cut can be read off whenever 0<ā_{k0}+γā_{i0}<1.

Formalization targets

Theorem 9.3 (goal) — the most-improving pivot column

The pivot column in row i most improving the cut from row k is indexed by l*∈J minimizing f⁺(γ_l) (if ā_{kl}ā_{il}<0) or f⁻(γ_l) (if ā_{kl}ā_{il}>0), over all l∈J with -ā_{k0}/ā_{i0}<γ_l<(1-ā_{k0})/ā_{i0}, γ_l:=-ā_{kl}/ā_{il}.

The chain of results building toward it

Lemma 9.1 (the tableau coefficients' closed form, eq. (9.4)-(9.5)) and Theorem 9.2 (the reduced costs of the CGLP columns u_i,v_i in terms of tableau coefficients, eq. (9.6)) are the two milestones the goal's own machinery is built from. Proposition 9.4, a bridge to Chapters 10-11's general split disjunctions, is included as a genuine milestone despite its payoff lying mostly outside this chapter.

Significance

The results themselves. This chapter is what makes lift-and-project cuts practical: instead of solving an (m+p+n)-row auxiliary LP from scratch for every candidate cut, a single pivot on the (LP)'s own tableau — guided by reduced costs that are themselves closed-form functions of tableau entries — identifies whether an improving cut exists and which one it is. Theorem 9.3's evaluation functions f⁺,f⁻ are exactly the tool a cutting-plane implementation would compute at every candidate pivot.

Formalizing it. No object in this mission exists on the platform prior to it or in Mathlib. This mission restates 08-cut-correspondence's (CGLP)_k apparatus locally, per the series convention and BRIEF.md's explicit instruction, and extends it with this chapter's own generalization to an arbitrary tableau row (needed since Lemma 9.1/Theorem 9.2 concern every basic variable's row, not only the disjunction row k).

Difficulty

Lemma 9.1's book proof is a four-case block-matrix verification (structural/surplus, basic/nonbasic); this mission instead states its content as the identity it is actually for — that the closed-form coefficients express every row's slack as an affine function of the nonbasic rows' slacks, for every point x — which follows tautologically from x=Â⁻¹b̂+Â⁻¹s_J's own definition once stated this way, without needing to reconstruct the block-matrix case analysis. Theorem 9.2's difficulty is that "reduced cost" is not already available as a formalized LP concept in this mission's apparatus; rather than build a generic LP reduced-cost theory, this mission follows the book's own derivation directly — explicitly constructing the pivoted-out extension of a basic solution (eq. (9.7)-(9.9)) and asserting that its objective value decomposes with r_{u_i},r_{v_i} as coefficients, which is genuine, non-circular content matching the proof's own final step ("we can then read the reduced costs... as the coefficients").

Formalization scope

This chapter makes the row/variable identification of Chapters 6-8 fully explicit (N directly indexes the structural variables), but no theorem's own displayed formula in this chunk needs that correspondence beyond what SurplusM's row-general treatment (this chunk's own generalization of 08-cut-correspondence's Surplus) already provides — see MODERATION_NOTES.md for why the S/N/B/R/P/Q block structure is proof machinery, not part of the stated content, throughout.

Theorem 9.3's range condition on γ_l, truncated in BRIEF.md's own excerpt, was completed by reading the PDF directly (confirmed identical to the range derived earlier in the same section): -ā_{k0}/ā_{i0}<γ_l<(1-ā_{k0})/ā_{i0}.

Selected references

• E. Balas, Disjunctive Programming, Springer, 2018. DOI: 10.1007/978-3-030-00148-3, Chapter 9.
• E. Balas, M. Perregaard, A precise correspondence between lift-and-project cuts, simple disjunctive cuts, and mixed integer Gomory cuts for 0-1 programming, Mathematical Programming B 94 (2003), 221–245 (cited in the text as [33], the origin of the tableau-pivoting procedure this chapter derives Lemma 9.1 and Theorem 9.2 from).

Motivation

Chapter 2's convex-hull machinery gives an exact, finitely-generated linear description of a disjunctive set's convex hull, but for a mixed 0-1 program with $p$ binary variables that description lives in a space with roughly $pn$ auxiliary variables — one full lift per disjunction. This chapter surveys the alternative nonlinear higher-dimensional constructions that several authors proposed for the same target, $\mathrm{conv}(K_0)$: multiplying the constraint system by products of $x_j$ and $1-x_j$ and linearizing the resulting quadratic terms, rather than disjoining and projecting one variable at a time. Two such constructions — Lovász and Schrijver's "cones of matrices" lift $N(K)$, and Sherali and Adams's hierarchy $K_t$ — both converge to the integer hull, and the chapter's central point is that both convergence proofs reduce, after all the nonlinear machinery is stripped away, to results already established by disjunctive programming's own one-variable-at-a-time convexification (Theorem 2.1, specialized here to Theorem 7.1, and the sequential-convexifiability theorem of Chapter 3).

Setting

$K := \{x \in \mathbb R^n : Ax \ge b,\ x \ge 0,\ x_j \le 1,\ j=1,\dots,p\} = \{x : \tilde A x \ge \tilde b\}$ is the LP relaxation of a mixed 0-1 program with $p$ of its $n$ variables 0-1 constrained, and $K_0 := K \cap \{x_j \in \{0,1\},\ j=1,\dots,p\}$ its feasible set. $P_j(K)$ (Section 7.1) multiplies $\tilde A x \ge \tilde b$ by $(1-x_j)$ and $x_j$, linearizes $y_i := x_ix_j$ and $x_j := x_j^2$, and projects onto $x$; iterating over a coordinate sequence gives $P_{i_1,\dots,i_t}(K)$. $N(K)$ (Section 7.2, Lovász-Schrijver) instead linearizes with a single symmetric matrix $Y$ ($Y_{ij} = Y_{ji} = x_ix_j$ for every pair) before projecting, and iterates as $N^t(K) := N(N^{t-1}(K))$. $K_t$ (Section 7.3, Sherali-Adams) multiplies by every product of $t$ literals $\prod_{j\in J_1}x_j\prod_{j\in J_2}(1-x_j)$ ($|J_1\cup J_2|=t$), linearizes each resulting monomial with a fresh "moment" variable, and projects.

Formalization targets

Theorem 7.6 (goal) — the Sherali-Adams hierarchy reaches the integer hull

$$K_p = \mathrm{conv}(K_0).$$

The chain of results building toward it

Theorem 7.1 ($P_j(K)$ equals the one-variable convex hull, a special case of Theorem 2.1), Theorem 7.2 (iterating $P_j$ over a fixed sequence reaches the hull of imposing $0/1$ on all of them), Corollary 7.3 (iterating over every 0-1 index reaches $\mathrm{conv}(K_0)$), Theorem 7.4 ($N(K) \subseteq P_j(K)$ for every $j$), Theorem 7.5 (iterating $N$ over $p$ steps reaches $\mathrm{conv}(K_0)$, by the same containment), and Theorem 7.7 ($K_t \subseteq P_{1,\dots,t}(K)$, proved by a genuine induction re-deriving every valid inequality of $P_{1,\dots,t}(K)$ from $(NL_t)$'s own rows).

Significance

The results themselves. This chapter is disjunctive programming's account of why three independently-developed convexification hierarchies — its own lift-and-project, Lovász-Schrijver, and Sherali-Adams — all reach the same integer hull: not by coincidence, but because each one's convergence proof is, at bottom, a disguised instance of the book's own Theorem 2.1 and Chapter 3 machinery. This is part of what situates disjunctive programming as the unifying framework behind several major lift-and-project hierarchies used throughout integer programming.

Formalizing it. No object in this mission exists on the platform prior to it or in Mathlib. This mission restates 03-sequential-convex's and 02a-convex-hull's vocabulary locally (per the series convention that a draft mission cannot import another draft mission's definitions), specialized throughout to the split disjunction $x_j \in \{0,1\}$.

Difficulty

The chapter's own account of the Sherali-Adams construction (Section 7.3) is narrative rather than displaying an explicit linear system for $(NL_t)$, unlike every other construction in this chapter (contrast eq. (7.1) and eq. (7.4), both displayed explicitly) — Step 1 says only "multiply $\tilde Ax \ge \tilde b$ with every product of the form $\prod_{j\in J_1}x_j \cdot \prod_{j\in J_2}(1-x_j)$." Deriving the actual linear system this multiplication produces requires expanding every $(1-x_j)$ factor via inclusion-exclusion over $S \subseteq J_2$ before the monomials can be linearized — a real, if mechanical, derivation step this mission had to carry out itself (RowNLt, see MODERATION_NOTES.md) rather than transcribe from a displayed equation. Theorem 7.7's own proof is a genuine argument (an induction re-deriving a valid inequality of $P_{1,\dots,t}(K)$ from $(NL_t)$'s rows layer by layer), not a restatement, so it is included as a milestone with real mathematical content rather than assumed.

Correcting BRIEF.md. The brief's "Recommended goal theorem" section mislabels Theorem 7.6 as the Lovász-Schrijver result "$K^p = \mathrm{conv}(K_0)$" — cross-checked directly against the PDF, Theorem 7.6 (p. 95, PDF 101) is stated [112] Kp = conv(K0), cited to Sherali and Adams, appearing immediately after Section 7.3 introduces $K_t$. The actual Lovász-Schrijver iteration-reaches-the-hull result, cited [99], is Theorem 7.5 (Np(K) = conv(K0)), formalized in this mission as a milestone (lovasz_schrijver_reaches_hull) rather than the goal. See STATUS.md.

Formalization scope

Three distinct convexification operators are kept fully distinguishable throughout, as BRIEF.md warns: Pj/IteratedSplit (one-variable-at-a-time, Section 7.1, Def_..._Basic), NOp/MK (Lovász-Schrijver, Section 7.2, Def_..._Lifts), and KtSet/IsXt (Sherali-Adams, Section 7.3, Def_..._Lifts) — no shared abbreviation or lemma conflates their defining predicates, even though all three converge to the same hull.

$N^t(K)$'s iteration (Theorem 7.5) is formalized via a dependent family of representations A t, b t for t : Fin (p+1) with a hypothesis relating consecutive steps to NOp's image at the previous step — the same pattern 04-normal-forms's Theorem 4.10 uses — since each successive $N^t(K)$ genuinely has a different, larger ambient constraint system, not a fixed matrix's power.

Theorem 7.7 is stated for an arbitrary $t$-subset $S \subseteq N'$ rather than the literal prefix $\{1,\dots,t\}$ the book's own statement uses (its own proof, by induction, treats an arbitrary inequality of $P_{1,\dots,t}(K)$ with no dependence on the prefix's specific ordering), which is what the goal theorem's proof, applied at $S = N'$, actually needs.

The Bienstock-Zuckerberg results quoted narratively in Section 7.5 ("Theorem 1"/"Theorem 2", [43]/[44]) are out-of-cone: the book itself presents them only as a survey of a further lift operator, under their own source papers' numbering, not as Balas's own numbered results, and does not restate their proofs.

Selected references

• E. Balas, Disjunctive Programming, Springer, 2018. DOI: 10.1007/978-3-030-00148-3, Chapter 7.
• L. Lovász, A. Schrijver, Cones of matrices and set-functions and 0-1 optimization, SIAM Journal on Optimization 1 (1991), 166-190 (cited in the text as [99], the origin of the $N(K)$ construction and Theorems 7.4-7.5).
• H.D. Sherali, W.P. Adams, A hierarchy of relaxations between the continuous and convex hull representations for zero-one programming problems, SIAM Journal on Discrete Mathematics 3 (1990), 411-430 (cited in the text as [112], the origin of the $K_t$ construction and Theorem 7.6).

Motivation

A logistic regression model is fitted on finitely many samples, and the quantity a practitioner cares about is the misclassification risk of the fitted classifier on new data. Its empirical counterpart, the training error, is biased downwards, and classical generalization bounds give it an additive margin that depends on a complexity measure of the model class rather than on the data at hand.

Shafieezadeh-Abadeh, Mohajerin Esfahani and Kuhn (NIPS 2015) take a distributionally robust route. They surround the empirical distribution of the training data by a ball of distributions in the Wasserstein metric and, for a given weight vector, compute the largest and the smallest misclassification probability over that ball. Their Theorem 3 shows that both extremes are optimal values of explicit linear programs. Combined with a measure-concentration result for the empirical distribution in the Wasserstein metric (Fournier and Guillin, PTRF 2015), the two values bracket the true risk with a prescribed confidence. The same Wasserstein-ball construction underlies the data-driven optimization framework of Mohajerin Esfahani and Kuhn (Math. Program. 2018).

Setting

Let $V$ be the feature space $\mathbb R^n$ with an arbitrary norm $\|\cdot\|$, and let labels take the values $y\in\{-1,+1\}$. The feature-label space is $\Xi = V\times\{-1,+1\}$ with points $\xi=(x,y)$. A weight vector $\beta$ acts on features by $x\mapsto\langle\beta,x\rangle$; its dual norm is $\|\beta\|_* = \sup_{\|x\|\le1}\langle\beta,x\rangle$.

Metric (Definition 2). For a weight $\kappa>0$,

$$d\big((x,y),(x',y')\big) = \|x-x'\| + \kappa\,|y-y'|/2 .$$

Changing a label costs $\kappa$; moving a feature costs its norm distance.

Wasserstein distance (Definition 1). For distributions $\mathbb Q,\mathbb P$ on $\Xi$, $W(\mathbb Q,\mathbb P)$ is the infimum of $\int d(\xi,\xi')\,\Pi(d\xi,d\xi')$ over all couplings $\Pi$ of $\mathbb Q$ and $\mathbb P$. The Wasserstein ball of radius $\varepsilon\ge0$ is $\mathbb B_\varepsilon(\mathbb P) = \{\mathbb Q : W(\mathbb Q,\mathbb P)\le\varepsilon\}$.

Data. Training samples $(\hat x_i,\hat y_i)$, $i=1,\dots,N$, define the empirical distribution $\hat{\mathbb P}_N = \frac1N\sum_{i=1}^N\delta_{(\hat x_i,\hat y_i)}$.

Classifier and risk. Logistic regression models $\operatorname{Prob}(y\mid x) = [1+\exp(-y\langle\beta,x\rangle)]^{-1}$ (eq. (1)). The classifier is $f_\beta(x)=+1$ if $\operatorname{Prob}(+1\mid x)>0.5$ and $-1$ otherwise, and its risk under the data-generating distribution $\mathbb P$ is $\mathfrak R(\beta) = \mathbb P[y\ne f_\beta(x)]$.

Worst- and best-case risks.

$$\mathfrak R_{\max}(\beta) = \sup_{\mathbb Q\in\mathbb B_\varepsilon(\hat{\mathbb P}_N)}\mathbb E^{\mathbb Q}\big[\mathbb 1_{\{y\langle\beta,x\rangle\le0\}}\big],\qquad \mathfrak R_{\min}(\beta) = \inf_{\mathbb Q\in\mathbb B_\varepsilon(\hat{\mathbb P}_N)}\mathbb E^{\mathbb Q}\big[\mathbb 1_{\{y\langle\beta,x\rangle<0\}}\big].$$

The worst case counts a nonpositive margin, the best case a strictly negative one.

The linear programs. For data $(\hat x_i,\hat y_i)$, a weight vector $\hat\beta$ and variables $\lambda\in\mathbb R$, $s,r,t\in\mathbb R^N$, program (10a) minimizes $\lambda\varepsilon + \frac1N\sum_i s_i$ subject to, for every $i$,

$$1 - r_i\hat y_i\langle\hat\beta,\hat x_i\rangle\le s_i,\quad 1 + t_i\hat y_i\langle\hat\beta,\hat x_i\rangle - \lambda\kappa\le s_i,\quad r_i\|\hat\beta\|_*\le\lambda,\quad t_i\|\hat\beta\|_*\le\lambda,\quad r_i,t_i,s_i\ge0 .$$

Program (10b) has the same objective and bounds, with the signs of the two margin terms exchanged.

Formalization targets

Goal: Theorem 3 (i)–(ii)

For every $\kappa>0$, $\varepsilon\ge0$, $N\ge1$, all samples and every weight vector $\hat\beta$, both programs attain their minima $v$ and $w$, and

$$\mathfrak R_{\max}(\hat\beta) = v,\qquad \mathfrak R_{\min}(\hat\beta) = 1-w .$$

The identities hold for each fixed $\hat\beta$, so they apply to any $\hat\beta$ computed from the data.

Milestone: Theorem 3(i) alone

$\mathfrak R_{\max}(\hat\beta)$ equals the minimum of (10a).

Milestones: the confidence clauses

If the training samples are i.i.d. from $\mathbb P$ and the radius is such that $\mathbb P^N\{\mathbb P\in\mathbb B_\varepsilon(\hat{\mathbb P}_N)\}\ge1-\eta$, then for any sample-dependent $\hat\beta$

$$\mathbb P^N\{\mathfrak R(\hat\beta)\le\mathfrak R_{\max}(\hat\beta)\}\ge1-\eta,\qquad \mathbb P^N\{\mathfrak R_{\min}(\hat\beta)\le\mathfrak R(\hat\beta)\}\ge1-\eta,$$ $$\mathbb P^N\{\mathfrak R_{\min}(\hat\beta)\le\mathfrak R(\hat\beta)\le\mathfrak R_{\max}(\hat\beta)\}\ge1-2\eta .$$

Significance

The result. Theorem 3 replaces an optimization over an infinite-dimensional set of distributions by a linear program with $3N+1$ variables and $4N$ constraints plus sign constraints. That makes the worst- and best-case misclassification probabilities computable at the scale of the training set, for any norm on the features whose dual norm can be evaluated. With the confidence clauses, the two values are data-driven upper and lower confidence bounds on the out-of-sample risk of the classifier actually deployed, including one fitted on the same data.

Formalizing it. The paper states Theorem 3 without proof in the main text; the argument is deferred to a technical appendix. No part of it is machine-checked. A formal proof needs the evaluation of a worst-case probability of a closed set over a type-1 Wasserstein ball around a discrete distribution, and the analogous best-case probability of an open set. Both are reusable in any Wasserstein-robust treatment of chance constraints or classification error.

Difficulty

The objective $\mathbf 1_{\{y\langle\beta,x\rangle\le0\}}$ is neither continuous nor concave, so the duality theorems for Wasserstein balls stated for continuous or Lipschitz losses do not apply directly. Upper semicontinuity of the indicator of a closed set is what matters, and the strict inequality in $\mathfrak R_{\min}$ has to be handled as the complement of a closed set. The transport cost couples a norm on the features with a discrete label-flip cost, so a sample can reach the misclassification region either by moving its feature to the hyperplane $\langle\hat\beta,x\rangle=0$ or by flipping its label, and the two options interact through the shared budget $\varepsilon$. Distances to the hyperplane are measured in the given norm and produce the dual norm $\|\hat\beta\|_*$. The degenerate weight $\hat\beta=0$ (every point on the hyperplane) must come out correctly without any division by $\|\hat\beta\|_*$.

Formalization scope

The feature space is a finite-dimensional real normed space V with an arbitrary norm, standing for $(\mathbb R^n,\|\cdot\|)$; the Euclidean norm is not assumed. A weight vector is a continuous linear functional V →L[ℝ] ℝ, and $\|\hat\beta\|_*$ is its operator norm, which is exactly the dual norm. Labels are Bool with an explicit embedding $\text{true}\mapsto+1$, $\text{false}\mapsto-1$; the metric of Definition 2 is written literally. The Wasserstein distance is ℝ≥0∞-valued, probabilities and expectations of indicators are measure values in $[0,\infty]$, and suprema and infima range exactly over the probability measures in the ball. "min" in (10a)/(10b) is formalized as attainment (IsLeast) of the objective over the feasible set. Samples are indexed by Fin N with $N\ge1$.

The following choices differ from a literal reading of the page:

• The paper says the risk "can be expressed as" $\mathbb E^{\mathbb P}[\mathbb 1_{\{y\langle\beta,x\rangle\le0\}}]$. This fails on the hyperplane $\langle\beta,x\rangle=0$, where $f_\beta(x)=-1$ is correct for $y=-1$. The mission defines $\mathfrak R(\beta)=\mathbb P[y\ne f_\beta(x)]$ from (1) and includes the true statement $\mathbb E^{\mathbb P}[\mathbb 1_{\{y\langle\beta,x\rangle<0\}}]\le\mathfrak R(\beta)\le\mathbb E^{\mathbb P}[\mathbb 1_{\{y\langle\beta,x\rangle\le0\}}]$ as a helper item.
• The choice $\varepsilon=\varepsilon_N(\eta)$ of (8) and the measure-concentration theorem behind it (Theorem 2) are not formalized. The confidence clauses take their conclusion, $\mathbb P^N\{\mathbb P\in\mathbb B_\varepsilon(\hat{\mathbb P}_N)\}\ge1-\eta$, as a hypothesis, and "with probability $1-\eta$" is read as "with probability at least $1-\eta$". The printed level $1-2\eta$ is kept for the two-sided bound.

Swapping the strict and non-strict inequalities in $\mathfrak R_{\max}$ and $\mathfrak R_{\min}$, restricting the supremum to measures supported on the sample points, or replacing the ball by a set that excludes non-discrete distributions would each change the theorem. None of these is an acceptable reformulation of the goal.

Useful infrastructure: couplings of a discrete measure with an arbitrary one, the distance from a point to a closed half-space in a general norm, and LP-duality arguments for fractional-knapsack-type programs. Proofs of the helper and confidence items, and any reusable lemma about worst-case probabilities of closed sets over Wasserstein balls, are welcome.

Selected references

• S. Shafieezadeh-Abadeh, P. Mohajerin Esfahani, D. Kuhn, Distributionally Robust Logistic Regression, Advances in Neural Information Processing Systems 28 (NIPS 2015). https://papers.nips.cc/paper/2015/hash/cc1aa436277138f61cda703991069eaf-Abstract.html
• N. Fournier, A. Guillin, On the rate of convergence in Wasserstein distance of the empirical measure, Probability Theory and Related Fields 162 (2015). https://doi.org/10.1007/s00440-014-0583-7
• P. Mohajerin Esfahani, D. Kuhn, Data-driven distributionally robust optimization using the Wasserstein metric: performance guarantees and tractable reformulations, Mathematical Programming 171 (2018). https://doi.org/10.1007/s10107-017-1172-1

Motivation

Many polytopes that arise from combinatorial optimization problems have no small facet description in their natural variable space, yet become describable by a compact linear system once lifted to a higher-dimensional space of auxiliary variables and projected back down — Chapter 2's own extended formulation of the convex hull of a disjunctive set is one instance of this phenomenon. This chapter turns the idea around: rather than using projection to build a compact formulation, it uses projection to prove integrality of a formulation that is already compact but whose integrality is not obvious from any standard sufficient condition (total unimodularity, balancedness, etc.). The technique is illustrated on three closely related combinatorial polytopes built from perfectly matchable, assignable, and path-decomposable vertex subsets of a graph or digraph — each proved integral by lifting to an edge- or arc-variable space where total unimodularity is easy to check, then projecting.

Setting

For a finite vertex set $V$, the incidence vector of $W \subseteq V$ is $1$ on $W$, $0$ elsewhere, and $x(S) := \sum_{i \in S} x_i$. A graph $G(W)$ has a perfect matching if there is a fixed-point-free involution on $W$ respecting adjacency. The PMS (Perfectly Matchable Subgraph) polytope of $G$ is $\mathrm{conv}(X)$ where $X$ is the set of incidence vectors of such $W$; $N(S) := \{j \notin S : (i,j) \in E \text{ for some } i \in S\}$.

For a digraph $(V,A)$: $G(W)$ is assignable if it admits a cycle decomposition (a permutation of $W$ respecting arcs), giving the Assignable Subgraph Polytope. For an acyclic digraph with distinguished nodes $s,t$: $G(W \cup \{s,t\})$ admits an $s$-$t$ path decomposition if a collection of interior-node-disjoint $s$-$t$ paths covers it, giving the $s$-$t$ Path Decomposable Subgraph Polytope over $W \subseteq V \setminus \{s,t\}$. $\Gamma(S)$ and $\Gamma^*(S)$ are the corresponding out-neighborhood operators. For an arbitrary graph, $c(S)$ counts the connected components of the induced subgraph $G(S)$.

Formalization targets

Theorem 5.1 (goal) — the PMS polytope of a bipartite graph

$$0 \le x_i \le 1\ (i \in V), \qquad x(V_1) - x(V_2) = 0, \qquad x(S) - x(N(S)) \le 0\ \ (S \subseteq V_1).$$

Theorem 5.2 — the Assignable Subgraph Polytope

$$0 \le x_i \le 1\ (i \in V), \qquad x(S \setminus \Gamma(S)) - x(\Gamma(S) \setminus S) \le 0 \quad (S \subseteq V).$$

Theorem 5.3 — the $s$-$t$ Path Decomposable Subgraph Polytope

$$0 \le x_i \le 1\ (i \in V), \qquad x(S \setminus \Gamma^*(S)) - x(\Gamma^*(S) \setminus S) \le 0 \quad (S \subseteq V \setminus \{s,t\}).$$

Theorem 5.4 — the PMS polytope of an arbitrary graph

$$0 \le x_i \le 1\ (i \in V), \qquad x(S) - x(N(S)) \le |S| - c(S)$$

for every $S$ all of whose components are single nodes or nonbipartite with odd order — the weakest faithful statement, since dropping the side condition would assert the inequality for subsets it does not hold for.

Significance

The results themselves. Each theorem gives an explicit, checkable linear system defining a polytope that arises naturally from a combinatorial covering/decomposition property, turning "does $G(W)$ have property $X$" into a linear-programming feasibility question. Theorem 5.1 is the one the book proves in full and the template for the other three: bipartite matching, digraph assignment, and acyclic-digraph path decomposition are structurally parallel problems (all reduce to checking a König–Hall-type combinatorial condition), and the same lift-and-project technique handles all three uniformly. Theorem 5.4 extends the idea to arbitrary (non-bipartite) graphs at the cost of a sharper right-hand side and a component-based side condition, connecting to Edmonds' classical matching-polytope theory while remaining a genuinely different object (a polytope of coverable vertex sets, not of matchings themselves).

Formalizing it. No object in this mission — the PMS, Assignable, or Path Decomposable Subgraph polytopes, or their defining neighbor operators — exists on the platform prior to this mission. The closest platform result, MetricTSP.pm_polytope_decomposition (Edmonds' perfect matching polytope theorem, in edge-variable space over a fixed vertex set requiring every vertex matched), is a genuinely different object from Theorem 5.4's PMS polytope (vertex-variable space, vertices may be left unmatched by design) and is not reused as a kind: reference item; it is noted here as related, not equivalent.

Difficulty

The natural first attempt tries to verify each polytope's integrality directly, by checking a known sufficient condition (total unimodularity, balancedness) on the displayed vertex-space system itself. This fails: the book states explicitly that (5.5)'s coefficient matrix is not totally unimodular, which is exactly why the lift-to-edge-variables step is necessary at all. The real content of each theorem is the two-part argument: (1) the lifted system in edge/arc variables is totally unimodular (checkable directly), so its polyhedron is integral; and (2) the vertex- space system is exactly the projection of the lifted one — a nontrivial fact requiring Chapter 2's projection machinery, not merely an unfolding of definitions. Theorem 5.4's extra difficulty, flagged explicitly in the text, is that its projection cone is not pointed, so the proof must work with a finite generating set rather than extreme rays, and it suffices to find a subset of generators producing every facet rather than a complete generating set — a genuinely harder argument the book itself outsources to a citation.

Formalization scope

Undirected graphs use Mathlib's SimpleGraph; digraphs use a bare relation A : V → V → Prop (not required symmetric or irreflexive, matching the book's unrestricted notion). Bipartition is recorded via part : V → Bool (decidable by construction) rather than two Set V halves, keeping the sums x(V_1), x(V_2) computable over Finsets throughout. IsAssignable uses Equiv.Perm on the vertex-set subtype, since a cycle decomposition is exactly a permutation. IsComponentOf and IsBipartiteOn (Theorem 5.4) are built directly from reachability and 2-colorability rather than Mathlib's induced-subgraph/ConnectedComponent API, matching the "maximal connected subset" reading of "component" the book's own prose intends.

IsPathDecomposable (Theorem 5.3) encodes "admits an $s$-$t$ path decomposition" via a degree-constrained arc set (every interior node has exactly one incoming and one outgoing chosen arc, none entering $s$ or leaving $t$, at least one leaving $s$) rather than an explicit list of vertex-disjoint paths — provably equivalent by the standard fact that an acyclic arc set with this degree pattern always decomposes into such a path family, and considerably lighter to state and reason about than constructing Path objects directly.

A trivializing formalization is ruled out explicitly: every theorem keeps the fractional box constraint $0 \le x_i \le 1$ rather than the integral $x_i \in \{0,1\}$ (per BRIEF.md's own warning, dropping the relaxation collapses the claim to a restatement of the combinatorial definition), and Theorem 5.1 is stated only for bipartite graphs — never generalized to subsume Theorem 5.4's genuinely different inequality system and side condition.

Selected references

• E. Balas, Disjunctive Programming, Springer, 2018. DOI: 10.1007/978-3-030-00148-3, Chapter 5, §5.2.
• M. O. Ball, U. Derigs, An analysis of alternate strategies for implementing matching algorithms, Networks 13 (1983) (cited in the text as [13], the origin of Theorems 5.2 and 5.3).
• W. R. Pulleyblank, J. Edmonds, Facets of 1-matching polyhedra, in Hypergraph Seminar, Springer Lecture Notes in Mathematics 411 (1974) — the origin of the perfectly matchable subgraph polytope literature (cited in the text as [34], the origin of Theorem 5.1).
• L. Lovász, M. D. Plummer, Matching Theory, Elsevier, 1986 (cited in the text as [35], the origin of Theorem 5.4).

Motivation

Convexity is what makes optimization tractable: a linear program's feasible region is convex, and this single fact underwrites the simplex method, LP duality, and everything built on top of them. Integer and disjunctive programs have no such luck — their feasible regions are unions of polyhedra, and a union of convex sets is generally not convex. If the convex hull of such a union could always be described compactly, integer programming would reduce to linear programming: optimize the same linear objective over the hull instead of the union, and any optimal vertex of the hull is automatically integral. The obstacle has always been that the convex hull of a union of polyhedra in $\mathbb{R}^n$, described directly by its facets in $\mathbb{R}^n$, typically needs exponentially many inequalities.

Balas's Theorem 2.1, proved in the 1970s and presented here as Chapter 2 of Disjunctive Programming (Balas, Springer 2018), breaks this exponential barrier by changing where the description lives. Rather than writing down the hull's facets in $\mathbb{R}^n$, Theorem 2.1 lifts the problem to a higher-dimensional space — one auxiliary copy of $\mathbb{R}^n$ per polyhedron in the union — where the hull becomes the projection of a single, explicitly given polyhedron whose size grows only linearly with the number of polyhedra. This "extended formulation" technique, born here, became one of the central tools of modern integer programming and combinatorial optimization: representing a hard polytope as the projection of an easy one in higher dimension underlies, for instance, the polynomial-size extended formulations known for many combinatorial polytopes.

Setting

Fix a finite index set $Q$. For $h \in Q$, let $A_h$ be a real matrix and $b_h$ a vector of matching row dimension, and set $P_h := \{x \in \mathbb{R}^n : A_h x \ge b_h\}$. The union $F := \bigcup_{h \in Q} P_h$ is the disjunctive set. Write $Q^* := \{h \in Q : P_h \ne \emptyset\}$ for the feasible disjuncts.

The recession cone of a nonempty polyhedron $P_h$ is $C_h := \{y : A_h y \ge 0\}$: the set of directions along which one can travel indefinitely from any point of $P_h$ while remaining in $P_h$. For a subset $M \subseteq Q$ and sets $S_h$ ($h \in M$), the (finite) Minkowski sum $\sum_{h \in M} S_h$ is $\{x : x = \sum_{h \in M} y^h \text{ for some } y^h \in S_h\}$. The maximal indices $Q^{**} \subseteq Q^*$ are the feasible disjuncts whose polyhedron is not contained in any other feasible disjunct's polyhedron.

Given a set $S \subseteq \mathbb{R}^n \times \beta$, its projection onto $x$ is $\mathrm{Proj}_x(S) := \{x : \exists\, y \in \beta,\ (x,y) \in S\}$.

Formalization targets

Theorem 2.1 (goal) — the convex hull of a disjunctive set

$$\mathrm{cl}\,\mathrm{conv}(F) = \mathrm{Proj}_x(P), \qquad P := \Big\{(x, \{y^h\}_{h \in Q^*}, \{y^h_0\}_{h \in Q^*}) : x = \!\!\sum_{h \in Q^*}\!\! y^h,\ A_h y^h - b_h y^h_0 \ge 0,\ y^h_0 \ge 0,\ \!\!\sum_{h \in Q^*}\!\! y^h_0 = 1 \Big\}.$$

This is the weakest correct statement: it claims only that the closed convex hull equals the projection of this specific lifted polyhedron $P$, not any stronger uniqueness or minimality claim about lifted representations in general (that refinement is Theorem 2.1's own follow-up discussion, not part of the theorem itself).

Corollary 2.2 — the extreme-point correspondence

Extreme points of $\mathrm{cl}\,\mathrm{conv}(F)$ correspond bijectively to the extreme points of $P$ that place all of their mass on a single disjunct's coordinates.

Theorem 2.3 — tightness of the lifted representation

$$P_Q = P \iff C_k \subseteq \sum_{h \in Q^*} C_h \quad \forall\, k \in Q \setminus Q^*,$$

where $P_Q$ is the variant of $P$ indexed by all of $Q$ rather than only $Q^*$.

Theorem 2.4 — from the convex hull to the union itself

Under two recession-cone conditions on $Q^{**}$, restricting $P_Q$'s $y^h_0$ variables to $\{0,1\}$ makes its $x$-projection recover $F$ itself, not merely $\mathrm{cl}\,\mathrm{conv}(F)$.

Significance

The result itself. Theorem 2.1 is the founding extended-formulation result of integer programming: it shows that every union of finitely many polyhedra — hence every mixed-integer program's feasible region, once expressed in disjunctive normal form — has a lifted description of size linear in the number of disjuncts, in stark contrast to the union's own facet description, which is generally exponential. Corollary 2.2 shows this lifting is not merely an upper bound with extraneous points: its extreme points correspond exactly, one-to-one, with the extreme points of the object it represents. Theorems 2.3 and 2.4 sharpen the picture: 2.3 tells you exactly when you can avoid knowing in advance which disjuncts are nonempty, and 2.4 tells you exactly when the same family of lifted systems, restricted to integral $y^h_0$, describes the union $F$ exactly rather than only its convex hull — this is Jeroslow and Lowe's characterization of when a disjunctive set is representable as the feasible region of an integer program at all.

Formalizing it. No object in this mission — the disjunctive set $F$, its lifted polyhedron $P$, recession cones of a union's components, or the extreme-point correspondence between a polytope and its lift — exists on the platform prior to this mission or anywhere in Mathlib (substrate.md records zero LP/polyhedron modules in Mathlib as of this writing). This mission is a from-scratch formalization of the book's central construction, restating (rather than importing) the disjunctive-set vocabulary introduced by the companion IntroDuality mission, per the series' convention that a draft mission cannot import another draft mission's definitions.

Difficulty

The natural first attempt at Theorem 2.1 tries to prove the two inclusions $\mathrm{cl}\,\mathrm{conv}(F) \subseteq \mathrm{Proj}_x(P)$ and $\mathrm{Proj}_x(P) \subseteq \mathrm{cl}\,\mathrm{conv}(F)$ by a direct facet-by-facet or vertex-by-vertex argument in $\mathbb{R}^n$ — exactly the exponential-size approach the theorem exists to avoid. The book's own first proof instead works entirely with convex combinations: an arbitrary point of $\mathrm{cl}\,\mathrm{conv}(F)$ is a combination of at most $|Q^*|$ points, one from each polyhedron in the union (Carathéodory-style), which converts directly into a point of $P$ by splitting the combination's weight across the lifted coordinates — and conversely, a point of $P$ decomposes, disjunct by disjunct, into a convex combination of that disjunct's own vertices and extreme rays. Neither direction ever needs to enumerate facets of $\mathrm{cl}\,\mathrm{conv}(F)$ in $\mathbb{R}^n$. The second proof (via projection and the polar cone $W$ of the lifted system) shows the projected inequalities coincide with exactly the valid-inequality characterization of Theorem 1.2 (disjunctive Farkas), which is a different, complementary way of seeing why no facet of $\mathrm{cl}\,\mathrm{conv}(F)$ is missed.

Formalization scope

All theorems are stated over a finite index set Q : Type* with [Fintype Q], matrices Matrix (Fin (m h)) (Fin n) ℝ with m : Q → ℕ allowed to depend on h, and vectors in Fin n → ℝ. DisjunctiveSet, FeasibleIndices, MaximalIndices, RecessionCone, and MinkowskiSumOver fix the chapter's vocabulary; ProjX, LiftedPolyhedron, and IntegerRestricted fix the lifted system and its variants. cl conv F is Mathlib's closure (convexHull ℝ ·); extreme points use Mathlib's Set.extremePoints.

LiftedPolyhedron ranges its auxiliary vectors $\{y^h\}$, $\{y^h_0\}$ over all of $Q$ rather than only the index subset Qidx the book restricts to, forcing the components outside Qidx to zero. This is an equivalent, Finset/decidability-free encoding — appending zero terms changes neither the defining sums nor the constraints — documented as a convention, not a weakening, in MODERATION_NOTES.md; the same definition instantiates both the $(2.1)$ system (Qidx = Q^*) and the $(2.1)_Q$ variant (Qidx = Q) that Theorem 2.3 compares.

A trivializing formalization is ruled out explicitly: taking $|Q^*| = 1$ collapses the lifted system to $x = y^1$, $y^1_0 = 1$, a vacuous restatement of $x \in P_1$ that proves nothing about unions. Every theorem here is stated for a generic finite Q, never specialized to a fixed small size. Contributions beyond this mission's statements would need genuine polyhedral machinery (vertex/extreme-ray decomposition of a polyhedron, Carathéodory's theorem for cones) that is itself absent from Mathlib and would be welcome as a separate, reusable definitions layer.

Selected references

• E. Balas, Disjunctive Programming, Springer, 2018. DOI: 10.1007/978-3-030-00148-3, Chapter 2, §2.1.
• E. Balas, Disjunctive programming: Properties of the convex hull of feasible points, Discrete Applied Mathematics 89 (1998), 3–44 (reprint of a 1974 MSRR, cited in the text as [6], the origin of Theorem 2.1).
• M. Conforti, M. Di Summa, Y. Faenza, On the size of extended formulations for polytopes associated with unions of polyhedra, SIAM Journal on Discrete Mathematics, cited in the text as [59] — establishes the tightness (minimum additional-variable count) of Theorem 2.1's lifted representation.
• R. G. Jeroslow, J. K. Lowe, Modelling with integer variables, Mathematical Programming Study 22 (1984), 167–184 (cited in the text as [86]; the characterization behind Theorem 2.4's significance).

Motivation

Linear programming duality is one of the load-bearing facts of optimization: every feasible linear program has a dual whose value matches the primal's, and this correspondence drives the simplex method's stopping criterion, sensitivity analysis, and most complexity results for polyhedral problems. Integer and mixed-integer programs have no such duality theorem in general — the feasible region of a mixed-integer program is not convex, and the entire apparatus of linear programming duality is built on convexity.

Disjunctive programming, introduced by Egon Balas in the early 1970s, closes part of this gap. A disjunctive set is a union of finitely many polyhedra rather than a single polyhedron — the natural convex-analytic shadow of the "either/or" logical structure that integer variables encode (an integer variable's feasible region is a finite union of half-open pieces, hence a disjunction of the linear constraints that pin it to each value). Balas's insight was that disjunctive programs — linear programs whose feasible region is such a union — admit a strong duality theorem of their own, generalizing the linear-programming case rather than replacing it. This mission formalizes that theorem (Theorem 1.5 of Balas, Disjunctive Programming, Springer 2018) together with the two results the same chapter builds around it: the founding construction of the field, the intersection cut (Theorem 1.1, circa 1970), and the disjunctive generalization of Farkas' Lemma (Theorem 1.2), which characterizes every valid inequality — hence every cutting plane — for a disjunctive set.

Setting

Fix a finite index set $Q$. For each $h \in Q$, let $A_h$ be a real $m_h \times n$ matrix and $b_h \in \mathbb{R}^{m_h}$, and set $P_h := \{x \in \mathbb{R}^n : A_h x \ge b_h\}$. The union $F := \bigcup_{h \in Q} P_h$ is a disjunctive set: any (linear) system of inequalities combined with the logical connectives "and", "or", "not" reduces, via its disjunctive normal form, to a set of exactly this shape. Because a union of convex sets need not be convex, $F$ is generally nonconvex even though each $P_h$ is a polyhedron.

A disjunctive program minimizes a linear objective over such a union:

$$(DP)\qquad z_0 = \min\Big\{ c x : x \in \textstyle\bigcup_{h \in Q} X_h \Big\}, \qquad X_h := \{x : A_h x \ge b_h,\ x \ge 0\}.$$

Its dual $(DD)$ pairs a scalar $w$ with one dual multiplier vector $u_h$ per disjunct, requiring $w \le u_h b_h$ and $u_h A_h \le c$, $u_h \ge 0$, simultaneously for every $h \in Q$, and maximizes $w$. Write $Q^* := \{h \in Q : X_h \ne \emptyset\}$ for the disjuncts whose primal system is feasible, and $Q^{**} := \{h \in Q : U_h \ne \emptyset\}$ (with $U_h := \{u_h \ge 0 : u_h A_h \le c\}$) for those whose dual system is feasible.

The theorems below also use two objects from the origin of the subject (§1.2): given a basic solution $\bar x$ of a linear program's optimal simplex tableau, with basic index set $I$ and nonbasic index set $J$, the tableau's coefficients $\bar a_{ij}$ ($i \in I$, $j \in J$) determine, for each nonbasic $j$, an extreme ray direction $r^j$ of the associated LP cone. A convex set $S$ is $P_I$-free at $\bar x$ if $\bar x$ lies in the interior of $S$ and that interior contains no point of the mixed-integer feasible set $P_I$.

Formalization targets

Theorem 1.1 — the intersection cut

$$\lambda^*_j := \max\{\lambda_j \ge 0 : \bar x + \lambda_j r^j \in S\}, \qquad \sum_{j \in J} \frac{1}{\lambda^*_j}\, x_j \ge 1.$$

The displayed inequality cuts off $\bar x$ but excludes no point of $P_I$, for any $P_I$-free convex set $S$ containing $\bar x$ in its interior.

Theorem 1.2 — Farkas' Lemma for Disjunctive Sets

$$\big(\forall x \in F,\ \alpha x \ge \alpha_0\big) \iff \big(\forall h \in Q^*,\ \exists\, u_h \ge 0,\ u_h A_h = \alpha,\ \alpha_0 \le u_h b_h\big).$$

Theorem 1.5 (goal) — duality for disjunctive programs

Under the Regularity Condition — $(Q^* \ne \emptyset$ and $Q \setminus Q^{**} \ne \emptyset) \Rightarrow Q^* \setminus Q^{**} \ne \emptyset$ — exactly one of:

1. both $(DP)$ and $(DD)$ are feasible, each attains an optimum, and $z_0 = w_0$; or
2. one of the two is infeasible, and the other is infeasible or has no finite optimum.

This is the weakest faithful statement of the theorem: it asserts only the shape of the dichotomy established by Balas, not any strengthened or specialized form of it.

Corollary 1.6 — necessity of the Regularity Condition

If the Regularity Condition fails, $(DP)$ is feasible, and $(DD)$ is infeasible, then $(DP)$ still has a finite minimum — exhibiting the duality gap that opens up once the condition is dropped.

Significance

The results themselves. Theorem 1.5 is the mission-critical fact that makes disjunctive programming a genuine extension of linear programming rather than an unrelated combinatorial device: every LP-duality-based algorithmic tool (bounding, sensitivity, complementary-slackness optimality certificates) has a disjunctive-programming counterpart because of this theorem. Theorem 1.1's intersection cut is the historical seed of an entire branch of integer-programming algorithms — lift-and-project cuts, mixed-integer Gomory cuts, and the split closure (later missions of this series) all specialize or generalize it. Theorem 1.2 is the structural fact that makes cutting-plane generation for disjunctive sets tractable at all: every valid inequality decomposes into per-disjunct Farkas certificates.

Formalizing it. None of these results, nor the union-of-polyhedra machinery they are stated over, exist on the platform prior to this mission: the platform's existing Farkas' Lemma and linear-programming strong duality theorems (SmaleNinth.farkas_lemma, SmaleNinth.lp_strong_duality) are the ordinary single-polyhedron statements, which is exactly the special case $|Q|=1$ of the theorems formalized here — genuinely different statements, not restatements. This mission is a from-scratch formalization of the disjunctive generalization, including the vocabulary (disjunctive sets, the paired primal/dual index sets $Q^*, Q^{**}$, the Regularity Condition) that the rest of the fifteen-mission Balas series builds on.

Difficulty

The obvious first attempt collapses the disjunctive dual $(DD)$ to $|Q|$ separate ordinary LP duals, one per disjunct, and tries to combine their individual strong-duality statements. This fails: $(DD)$ couples all disjuncts through the single shared scalar $w$, which must simultaneously satisfy $w \le u_h b_h$ for every $h \in Q$ at once, not disjunct-by-disjunct. The Regularity Condition exists precisely because this coupling can break down — Balas's own example (a two-term disjunctive program with an infeasible dual but a feasible, bounded primal) shows that without the condition, situation (2) of the dichotomy can fail: the primal can have a finite optimum with no matching dual optimum. Any formalization that omits the Regularity Condition, or weakens it to an informal restriction like "nondegenerate", either proves a false statement or proves nothing (a vacuous hypothesis), which Corollary 1.6 exists specifically to rule out.

Formalization scope

All three theorems are stated over a finite index set Q : Type* with [Fintype Q], real matrices Matrix (Fin (m h)) (Fin n) ℝ with row-dimension m : Q → ℕ allowed to depend on h (the book never assumes a common row count across disjuncts), and vectors in Fin n → ℝ. Poly, PolyNonneg, and DualPoly are the plain, nonnegative-orthant, and dual polyhedral systems respectively; FeasibleIndices and RegularityCondition pin $Q^*$/$Q^{**}$ and the Regularity Condition exactly as stated on p. 13. "No finite optimum" is formalized via UnboundedBelowOn / UnboundedAboveOn: nonempty (feasible) together with no finite bound on the objective, matching Balas's case (2), which explicitly distinguishes infeasibility from unboundedness.

A trivializing formalization is ruled out explicitly: fixing $|Q| = 1$ collapses Theorem 1.5 to ordinary LP duality (already on the platform) and Theorem 1.2 to ordinary Farkas' Lemma, so both theorems are stated for a generic finite Q, never specialized. Theorem 1.5's "exactly one of" dichotomy is formalized as a logical Xor of the two situations, not a weaker Or, since the book asserts mutual exclusivity, not merely that one holds.

The intersection-cut theorem (1.1) is formalized over a generic finite index type ι standing for the full set of structural and surplus variables, with I J : Finset ι the basic/nonbasic partition; a complete development would additionally need the simplex-tableau apparatus connecting ι, I, J, and abar to an actual linear program, which lies outside this mission and belongs instead to the tableau-focused later missions of the series (SimplexTableau, RayCGLP). The extremeRay and PIFree definitions introduced here are local to this mission and are restated, not imported, by later missions that need related vocabulary — per the series' convention that a draft mission cannot import another draft mission's definitions.

Selected references

• E. Balas, Disjunctive Programming, Springer, 2018. DOI: 10.1007/978-3-030-00148-3, Chapter 1.
• E. Balas, Intersection cuts — a new type of cutting planes for integer programming, Operations Research 19 (1971), 19–39. (Theorem 1.1's origin, cited in the text as [4].)
• E. Balas, Disjunctive programming, Annals of Discrete Mathematics 5 (1979), 3–51. (Cited in the text as [9], the origin of Theorem 1.5.)

Motivation

Subgradient optimization maximizes a concave function that is not differentiable by stepping along an arbitrary subgradient with a prescribed sequence of step sizes. It became a standard tool of integer programming after Held and Karp used it to compute the Lagrangian 1-tree bound for the traveling-salesman problem (Held & Karp 1971). Held, Wolfe and Crowder then tested it on the assignment problem, a traveling-salesman relaxation and a multicommodity flow problem (Held, Wolfe & Crowder 1974).

The method has one practical defect that the paper names at the start of its Section 6: it contains no test of optimality. The value $w(\pi^j)$ approaches the maximum, but at no finite step does the method say that the maximum has been reached, or what the maximum is. Section 6 of the paper supplies such a test for the case where $w$ is a minimum of finitely many affine functions. The finitely many subgradients produced by the iterates define a small linear program, the core problem, and from some iteration on this linear program already solves the full dual linear program. Its optimal value is therefore the exact maximum of $w$, obtained from quantities the method computes anyway. This is how the authors certified the optimal values reported in their experiments.

Timeline:

• 1967–1969: Poljak proves that the subgradient iterates satisfy $w(\pi^j)\to\max w$ when the step sizes tend to zero and have divergent sum (Poljak 1967; Poljak 1969).
• 1971: Held and Karp apply the method to the 1-tree bound (Held & Karp 1971).
• 1974: Held, Wolfe and Crowder prove that the core problem $P(J,J^*)$ solves the dual linear program (Theorem 6.3) and give a sufficient condition for bounded iterates (Theorem 6.1).
• 1996–1999: primal recovery from subgradient iterates is developed further, by convex combinations of the subgradients with weights derived from the step sizes (Sherali & Choi 1996; Larsson, Patriksson & Strömberg 1999).

Setting

Fix $n\ge0$ and write $E^n=\mathbb R^n$ with the Euclidean inner product $\pi\cdot v$. The data are $K\ge1$ scalars $c_k$ and vectors $v_k\in E^n$, and

$$w(\pi)=\min\{c_k+\pi\cdot v_k : k=1,\dots,K\}.\qquad(2.2)$$

The function $w$ is assumed bounded above, the paper's standing assumption. An index $k$ attains the minimum at $\pi$ if $c_k+\pi\cdot v_k=w(\pi)$.

A run of the subgradient algorithm consists of a starting point $\pi^0\in E^n$, step sizes $t_j>0$ and indices $k(j)$ such that $k(j)$ attains the minimum at $\pi^j$, and

$$\pi^{j+1}=\pi^j+t_j\,v_{k(j)}\qquad(j=0,1,\dots).\qquad(2.6)$$

No rule for choosing among several minimizing indices is imposed. Write $v^j=v_{k(j)}$ and $c^j=c_{k(j)}$. The step-size conditions are

$$t_j\to0,\qquad \sum_{j=0}^\infty t_j=\infty.\qquad(2.7)$$

The dual linear program of $\max w$ is

$$\min\Big\{\sum_k c_ky_k : y_k\ge0,\ \sum_ky_k=1,\ \sum_ky_kv_k=0\Big\}.\qquad(6.1)$$

For integers $J<J^*$ the core problem $P(J,J^*)$ has one variable $y_j$ for each iteration $j\in[J,J^*]$:

$$\min\Big\{\sum_{j=J}^{J^*}c^jy_j : y_j\ge0,\ \sum_{j=J}^{J^*}y_j=1,\ \sum_{j=J}^{J^*}y_jv^j=0\Big\}.$$

An index chosen at several iterations contributes several identical columns. A point $y$ of $P(J,J^*)$ is sent to the point $\bar y_k=\sum\{y_j : J\le j\le J^*,\ k(j)=k\}$ of (6.1). This aggregation preserves feasibility and objective value.

Formalization targets

Goal: Theorem 6.3 (p. 82)

Assume $w$ is bounded above, $(t_j,\pi^j,k(j))$ is a run satisfying (2.7), and $\{\pi^j\}$ is bounded. Then

$$\forall J\ \exists J^*>J:\quad P(J,J^*)\text{ has a solution, and every solution of }P(J,J^*)\text{ aggregates to a solution of (6.1)}.$$

The goal states existence of $J^*$, which is what the paper claims. The paper's argument in fact gives the conclusion for every sufficiently large $J^*$. That stronger form is not the goal. Feasibility of $P(J,J^*)$ (Lemma 6.2) or the inequality $\mathrm{Value}[P(J,J^*)]\ge\mathrm{Value}[(6.1)]$, which holds for every feasible $P(J,J^*)$, is not a formalization of the goal. The content is optimality in (6.1).

Milestones

1. Eq. (2.10): if $\pi^*$ maximizes $w$ and $k$ attains the minimum at $\pi$, then $w^*-w(\pi)\le v_k\cdot(\pi^*-\pi)$.
2. §6, p. 80 (display): under (2.6), (2.7) and $w$ bounded above, $\lim_j w(\pi^j)=\max w=w(\pi^*)$ for some $\pi^*$. The iterates are not assumed bounded.
3. Theorem 6.1: if every $\pi\ne0$ has some $\pi\cdot v_k<0$, every run satisfying (2.7) is bounded.
4. Eq. (6.1): (6.1) has a solution, and its optimal value equals $\max w$.
5. Lemma 6.2: for any $J$ there is $J^*>J$ with $P(J,J^*)$ feasible, for bounded runs.

Significance

Theorem 6.3 turns an asymptotic method into one that returns an exact answer. Solving $P(J,J^*)$ for growing $J^*$ produces a linear program of bounded size whose optimum is eventually the optimum of (6.1), and hence $\max w$. In the Lagrangian applications, where (6.1) is the linear relaxation of a combinatorial problem, this yields both the bound and a primal solution of the relaxation. The theorem is the ancestor of the primal-recovery results listed in the timeline.

The mission produces a machine-checked version of the paper's Section 6, together with the input the paper takes on citation: Poljak's convergence theorem for divergent-series step sizes, specialized to piecewise-linear concave functions. Neither Poljak's theorem nor Theorem 6.3 is in Mathlib. The pieces are reusable: the convergence theorem applies to every Lagrangian dual solved by subgradient steps, and the duality between $\max w$ and (6.1) is linear-programming duality for a minimum of affine functions.

Difficulty

The inequality $\operatorname{Value}P(J,J^*)\ge\operatorname{Value}(6.1)$ is immediate, since aggregation maps feasible points to feasible points with the same objective. All of the content lies in the reverse inequality. That inequality ties a finite linear program to the limit of an infinite sequence, and it must hold for an arbitrary choice among tied minimizing indices. The iterates themselves need not converge, and under (2.7) the values $w(\pi^j)$ are not monotone. So an argument that inspects a single iterate, or assumes that the method settles on one face of $w$, fails. The convergence statement of milestone 2 is not proved in the paper and is the heaviest single step. Feasibility of $P(J,J^*)$ also needs its own argument, and it fails without the boundedness hypothesis.

Formalization scope

$E^n$ is EuclideanSpace ℝ (Fin n), the index set is a finite nonempty type ι, and $w$ is the finite minimum Finset.univ.inf'. A run is the predicate IsSubgradientRun c v t π k: positive steps, a minimizing index at every step, and update (2.6). It is not a function of $\pi^0$, so every tie-breaking rule is covered. (2.7) is StepSizeCond t: t → 0, and the partial sums tend to $+\infty$. Iterates are indexed from $j=0$. Boundedness is Bornology.IsBounded (Set.range π). The variables of $P(J,J^*)$ are a function on $\mathbb N$ of which only the values at $J\le j\le J^*$ enter. Optimality of $y$ in either linear program means feasibility plus an objective no larger than that of every feasible point. Suprema are never taken over unbounded sets: every maximum of $w$ is stated as attained at an explicit $\pi^*$.

A statement that only asserts feasibility of $P(J,J^*)$, or only $\operatorname{Value}P\ge\operatorname{Value}(6.1)$, is not the theorem. The goal requires that the solutions of $P(J,J^*)$ be optimal for (6.1).

Theorem 6.1 is printed for the step rule (2.8), but its proof uses $w(\pi^j)\to w^*$, the consequence of (2.7). The mission states it for (2.7), and its milestone title says so.

A complete development needs:

• linear-programming duality for (6.1), including attainment;
• the convergence theorem for divergent-series step sizes;
• existence of a maximizer of a bounded-above minimum of finitely many affine functions;
• basic facts on convex hulls of finitely many vectors in $E^n$.

The first three are reusable well beyond this mission. Contributions of any of them, as standalone theorems, are welcome.

Selected references

• M. Held, P. Wolfe, H. P. Crowder, Validation of subgradient optimization, Mathematical Programming 6 (1974) 62–88. https://doi.org/10.1007/BF01580223
• M. Held, R. M. Karp, The traveling-salesman problem and minimum spanning trees: Part II, Mathematical Programming 1 (1971) 6–25. https://doi.org/10.1007/BF01584070
• B. T. Poljak, A general method of solving extremum problems, Soviet Mathematics Doklady 8 (1967) 593–597.
• B. T. Poljak, Minimization of unsmooth functionals, USSR Computational Mathematics and Mathematical Physics 9 (1969) 14–29. https://doi.org/10.1016/0041-5553(69)90061-5
• H. D. Sherali, G. Choi, Recovery of primal solutions when using subgradient optimization methods to solve Lagrangian duals of linear programs, Operations Research Letters 19 (1996) 105–113. https://doi.org/10.1016/0167-6377(96)00019-3
• T. Larsson, M. Patriksson, A.-B. Strömberg, Ergodic, primal convergence in dual subgradient schemes for convex programming, Mathematical Programming 86 (1999) 283–312. https://doi.org/10.1007/s101070050090