Fact-checked by Grok 2 weeks ago

Dual linear program

In linear programming, the dual linear program (or simply the dual) is a canonical optimization problem derived from a given primal linear program, where the roles of the objective function and constraints are interchanged to provide complementary insights into the optimization landscape. For a primal problem formulated as maximizing c^T x subject to A x \leq b and x \geq 0, the dual is the minimization problem \min b^T y subject to A^T y \geq c and y \geq 0, with one dual variable y_i corresponding to each primal inequality constraint and one dual constraint per primal variable. This construction ensures that the dual problem yields an upper bound on the primal's optimal objective value for any feasible dual solution. The foundational weak duality theorem states that if both the and problems are feasible, then the optimal value of the maximization is less than or equal to the optimal value of the minimization, reflecting the bounding relationship between their objective functions. Under the strong duality theorem, if either problem is feasible and bounded, then both attain optimal solutions with equal objective values, establishing a profound equivalence that links the two formulations. Additionally, the dual of the dual recovers the original , underscoring the of the duality framework. Duality in linear programming extends beyond theoretical symmetry to practical interpretations, where dual variables represent shadow prices or marginal values of the primal's resources, aiding in applications such as and economic modeling. The complementary slackness theorem further connects optimal and solutions, stipulating that for each constraint, either it holds with equality in the primal or the corresponding variable is zero, and vice versa, which facilitates verification of optimality and economic interpretations. These principles, rooted in mid-20th-century developments in optimization theory, underpin efficient algorithms like the simplex method and enable formulations to simplify computation when the primal has many constraints.

Primal and Dual Problem Formulations

Primal Linear Program

The linear program in standard inequality form is defined as the of maximizing a linear subject to linear constraints and non-negativity conditions on the variables. Formally, it is expressed as \max \, c^\top x \quad \subjectto \, Ax \leq b, \quad x \geq 0, where x \in \mathbb{R}^n represents the of decision variables, c \in \mathbb{R}^n is the of coefficients for the , A \in \mathbb{R}^{m \times n} is the matrix of constraint coefficients, and b \in \mathbb{R}^m is the specifying the right-hand side values of the constraints. In this formulation, the decision variables x quantify the levels of activities or allocations being optimized, while the objective coefficients c reflect the contribution or value of each unit of those activities toward the goal of maximization. The constraints Ax \leq b model limitations such as available resources or capacities, with each row of A and corresponding entry in b defining a single resource bound. Linear programming, and thus the primal problem, originated as a tool for solving optimization challenges in contexts like , where finite supplies must be distributed across competing demands to achieve maximal or . The formulation assumes a finite-dimensional setting with all coefficients and variables being real numbers, ensuring the problem is well-defined over the .

Dual Linear Program

The dual linear program is derived from a given primal linear program in standard form, which seeks to maximize \mathbf{c}^T \mathbf{x} subject to A \mathbf{x} \leq \mathbf{b} and \mathbf{x} \geq \mathbf{0}, where A is an m \times n matrix, \mathbf{b} \in \mathbb{R}^m, and \mathbf{c} \in \mathbb{R}^n. The standard dual formulation then takes the form of a minimization problem: minimize \mathbf{b}^T \mathbf{y} subject to A^T \mathbf{y} \geq \mathbf{c} and \mathbf{y} \geq \mathbf{0}, where \mathbf{y} \in \mathbb{R}^m represents the dual variables and A^T denotes the transpose of the primal constraint matrix A. This structure ensures that the dual constraints correspond to the primal variables, with the transposed matrix A^T linking the two problems while preserving the non-negativity requirements. The dual variables \mathbf{y} serve as shadow prices or Lagrange multipliers associated with the primal constraints, quantifying the marginal value of relaxing each resource limit in the original problem. In this role, each component y_i indicates the increase in the objective value per unit increase in the right-hand side b_i of the corresponding constraint, assuming optimality. This interpretation highlights the dual's focus on efficiency, where \mathbf{y} scales the primal inequalities to bound the from above. In terms of optimization, the dual problem is equivalent to the primal in the sense that it converts the maximization objective into a minimization while maintaining the same optimal value under feasibility and boundedness conditions. Specifically, the dual objective \mathbf{b}^T \mathbf{y} provides an upper bound on the primal maximum \mathbf{c}^T \mathbf{x} for any feasible solutions, reflecting the inherent symmetry between the two formulations. This min-max relationship underscores the dual's utility in verifying primal solutions and exploring alternative optimization perspectives.

Interpretation of Dual Variables

In linear programming, the dual variables y_i, corresponding to the inequality constraints in the primal problem, are interpreted as shadow prices, which quantify the marginal increase in the optimal value of the primal objective function resulting from a unit increase in the right-hand side coefficient of the i-th primal constraint. This interpretation stems from the economic perspective where these variables represent the imputed value or of scarce limited by the primal constraints. For instance, in a profit-maximization primal problem subject to resource limitations, a positive shadow price y_i indicates the additional profit attainable if one more unit of the i-th were available, guiding decisions on resource acquisition or expansion. Geometrically, the vector of optimal dual variables y^* defines the normal to a that touches the primal feasible set—a polyhedron—at an optimal primal point x^*, without penetrating the interior of the set. This separates the from points where the primal objective would be worse, ensuring that the dual objective b^T y^* provides a tight lower bound equal to the primal optimum under . The components y_i^* weight the contributions of individual to this supporting structure, reflecting how tightly each binds at optimality; zero values for y_i^* correspond to non-binding constraints where the does not touch the defined by that inequality. In contexts for maximization primal problems, the shadow prices y_i embody the per unit of each constrained resource, enabling the total imputed of all resources to equal the optimal . Nonzero shadow prices signal fully utilized resources, while their magnitudes inform trade-offs, such as whether relaxing a justifies additional costs. This interpretation facilitates , revealing how perturbations in resource availability affect overall optimality without resolving the entire problem. Under non-degeneracy of the optimal solution—where no is redundant at the —the optimal variables and associated prices are unique, providing a precise marginal valuation for each . Non-degeneracy ensures a single aligns exactly with the binding constraints at x^*, avoiding ambiguity in the geometric or economic interpretations. In contrast, degeneracy can yield multiple optima, but the standard interpretation assumes this non-degenerate case for clarity and applicability in most theoretical analyses.

Constructing the Dual Program

Rules for Standard Forms

The standard form of a linear program is formulated as maximizing the objective c^T x subject to inequality constraints Ax \leq b and non-negativity constraints x \geq 0, where A is the , b is the right-hand-side , c is the objective , x is the variable , and all components are appropriately dimensioned. The corresponding dual linear program is then constructed by systematically transforming these elements according to established primal-dual correspondences. The construction follows three primary rules for standard inequality forms:
  1. The primal maximization transforms into a minimization , and the primal constraints (Ax \leq b) become non-negative variables (y \geq 0), with the coefficients taken from the primal right-hand-side vector b. Thus, the is \min b^T y.
  2. Each primal variable x_j \geq 0 corresponds to a dual inequality constraint, where the dual constraint coefficients are derived from the transpose of the primal coefficient matrix A^T, the right-hand side is the primal objective coefficient c_j, and the inequality direction is \geq. Thus, the dual constraints are A^T y \geq c.
  3. The number of dual variables equals the number of primal inequality constraints (excluding non-negativity), and the number of dual constraints equals the number of primal variables. The dual variables y_i are unrestricted in sign except for non-negativity (y \geq 0) to match the primal's \leq constraints.
These rules ensure a direct mapping between primal and dual components, as summarized in the following table of primal-dual pairings:
Primal ElementDual Element
Maximization c^T xMinimization b^T y
Inequality constraints Ax \leq bNon-negative variables y \geq 0
Non-negative variables x_j \geq 0Inequality constraints A^T y \geq c
Right-hand side b coefficients in b^T y
coefficients cRight-hand side in A^T y \geq c

Handling General Constraints

In linear programming, primal problems often include equality constraints, free (unrestricted) variables, or variables with mixed sign restrictions beyond non-negativity, requiring extensions to the standard dual construction rules. To handle equality constraints in the primal, such as \sum_j a_{ij} x_j = b_i, one approach is to convert the equality into two inequalities: \sum_j a_{ij} x_j \leq b_i and \sum_j a_{ij} x_j \geq b_i. This introduces two dual variables, one unrestricted in sign for each inequality, which can be combined to yield an effectively unrestricted dual variable y_i for the original equality. Alternatively, the dual can be formed directly by associating an unrestricted dual variable y_i \in \mathbb{R} with the equality constraint, without splitting, as the Lagrangian multiplier for an equality lacks sign restrictions. Free variables in the , where x_j \in \mathbb{R} without sign bounds, correspond to constraints in the formulation. Specifically, for a free variable x_j, the associated constraint becomes \sum_i a_{ij} y_i = c_j, reflecting the absence of sign restrictions on x_j that would otherwise impose directions. For variables that are non-positive (x_j \leq 0) rather than non-negative, the constraints undergo a sign flip compared to the standard case. In a maximization , a non-positive x_j leads to a constraint of the form \sum_i a_{ij} y_i \leq c_j, whereas a non-negative x_j yields \sum_i a_{ij} y_i \geq c_j; this ensures the respects the variable's bounded direction. Similarly, for non-positive variables arising from certain constraints, the signs adjust accordingly to maintain duality consistency. To construct the dual for such general primal forms, a canonical transformation procedure first standardizes the problem before applying the base rules for inequality constraints and non-negative variables. This involves: (1) replacing equalities with pairs of inequalities and introducing or surplus variables as needed; (2) substituting variables x_j with differences of non-negative variables (x_j = x_j^+ - x_j^-, x_j^+, x_j^- \geq 0); (3) flipping signs for non-positive variables by setting x_j = -z_j with z_j \geq 0 and adjusting coefficients; and (4) converting the objective if necessary (e.g., maximization to minimization via ). Once in standard form—all constraints as \leq inequalities and variables \geq 0—the standard rules can be applied directly.
Primal Constraint TypeDual Variable SignPrimal Variable TypeDual Constraint Type
\sum_j a_{ij} x_j \leq b_i (max primal)y_i \geq 0x_j \geq 0\sum_i a_{ij} y_i \geq c_j
\sum_j a_{ij} x_j = b_iy_i unrestrictedx_j free\sum_i a_{ij} y_i = c_j
\sum_j a_{ij} x_j \geq b_i (max primal)y_i \leq 0x_j \leq 0\sum_i a_{ij} y_i \leq c_j
This table summarizes the sign correspondences for a maximization primal, building on the standard rules for non-negative variables and \leq constraints.

Duality Theorems

Weak Duality Theorem

The weak duality theorem establishes a fundamental inequality between the objective values of a primal linear program and its dual for any pair of feasible solutions. Consider the standard primal maximization problem: maximize \mathbf{c}^T \mathbf{x} subject to A \mathbf{x} \leq \mathbf{b} and \mathbf{x} \geq \mathbf{0}, with feasible set \mathcal{P}, and its corresponding dual minimization problem: minimize \mathbf{b}^T \mathbf{y} subject to A^T \mathbf{y} \geq \mathbf{c} and \mathbf{y} \geq \mathbf{0}, with feasible set \mathcal{D}. For any \mathbf{x} \in \mathcal{P} and \mathbf{y} \in \mathcal{D}, the theorem states that \mathbf{c}^T \mathbf{x} \leq \mathbf{b}^T \mathbf{y}. The proof relies on the non-negativity of the dual variables and the primal constraints. Starting from the dual constraint, A^T \mathbf{y} \geq \mathbf{c}, transpose and multiply by the primal feasible \mathbf{x} \geq \mathbf{0} to obtain \mathbf{c}^T \mathbf{x} \leq (A^T \mathbf{y})^T \mathbf{x} = \mathbf{y}^T (A \mathbf{x}). Since \mathbf{x} \in \mathcal{P} implies A \mathbf{x} \leq \mathbf{b} and \mathbf{y} \geq \mathbf{0}, it follows that \mathbf{y}^T (A \mathbf{x}) \leq \mathbf{y}^T \mathbf{b}. Combining these yields \mathbf{c}^T \mathbf{x} \leq \mathbf{b}^T \mathbf{y}. This inequality implies that any feasible dual solution provides an upper bound on the optimal value of the problem, assuming the is feasible and bounded. If the optimum is finite, the must also be feasible and its objective value matches or exceeds it, serving as a practical tool for bounding optimization problems without solving them fully. No assumptions of finiteness, attainment of optima, or additional duality conditions are required for the to hold.

Strong Duality Theorem

The strong duality theorem in asserts that if the problem attains a finite optimal value, then the problem also attains a finite optimal value, and these optimal values are equal. For the standard maximization problem \max c^\top x subject to Ax \leq b and x \geq 0, and its corresponding minimization problem \min b^\top y subject to A^\top y \geq c and y \geq 0, the guarantees p^* = d^*, where p^* and d^* denote the respective optimal objective values, provided the is feasible and bounded. This equality holds without additional qualification constraints like , as linear programs are polyhedral and follows directly from feasibility and boundedness of one problem implying the same for the . The conditions for in require that at least one of the problems ( or ) is feasible and has a finite optimum; the other then inherits these properties with equal objective value. In the polyhedral setting of linear programs, this is ensured by the absence of infeasibility or unboundedness in both simultaneously when one is solvable. Building on the weak duality inequality p^* \leq d^*, closes the gap to equality under these feasibility assumptions. A proof sketch via the simplex method highlights attainment of equality: an optimal to the primal yields shadow prices (dual variables) that satisfy dual feasibility, with the primal and dual objectives matching due to zero reduced costs at optimality. Alternatively, a separating argument in the feasible set confirms that no exists when optima are finite, as any separating plane would contradict boundedness or feasibility. Thus, the —defined as d^* - p^*—is zero whenever applies.

Complementary Slackness

Complementary slackness conditions provide a precise of the optimal solutions to a -dual pair of linear programs. Specifically, if x^* is an optimal solution to the problem \max \{ c^T x \mid A x \leq b, x \geq 0 \} and y^* is an optimal solution to the dual problem \min \{ b^T y \mid A^T y \geq c, y \geq 0 \}, then the following hold for all indices i = 1, \dots, m and j = 1, \dots, n: y_i^* (b_i - (A x^*)_i) = 0, \quad x_j^* ((A^T y^*)_j - c_j) = 0. These conditions ensure that the products of corresponding primal and dual slacks and variables are zero at optimality. The interpretation of these conditions is that, at the optimum, for each primal constraint i, either the dual variable y_i^* is zero (indicating the constraint does not bind the dual solution) or the primal slack b_i - (A x^*)_i is zero (meaning the constraint is tight). Symmetrically, for each dual constraint j, either the primal variable x_j^* is zero or the dual slack (A^T y^*)_j - c_j is zero. This mutual exclusivity highlights how optimal solutions "complement" each other by activating only when the opposing slack is absent, reflecting the economic intuition that shadow prices (dual variables) are positive only for binding resource constraints in the primal. The proof of complementary slackness follows from the theorem, which establishes that the and optimal values are equal, combined with the weak . Under weak duality, the difference between dual and primal objectives is b^T y - c^T x = \sum_i y_i (b_i - (A x)_i) + \sum_j x_j ((A^T y)_j - c_j) \geq 0 for feasible x, y. At optimality, this difference is zero, so each nonnegative term in the sums must individually be zero, yielding the slackness products. enables this tightening by guaranteeing the objectives coincide. These conditions play a crucial role in verifying optimality: if feasible solutions x and y to the and , respectively, satisfy the complementary slackness equalities, then both are optimal, as the weak closes to zero. Conversely, all optimal pairs satisfy them, providing a practical test for solution correctness without resolving the full program.

Mathematical Formulations

Vector Representations

The vector representation of a linear program provides a compact that emphasizes the inner product structure of the objective and constraints, facilitating analysis in finite-dimensional spaces. The problem in standard form is expressed as maximizing the \mathbf{c} \cdot \mathbf{x}, where \mathbf{c} \in \mathbb{R}^n is the coefficient and \mathbf{x} \in \mathbb{R}^n is the decision variable , subject to the constraints A \mathbf{x} \leq \mathbf{b} with A \in \mathbb{R}^{m \times n} the and \mathbf{b} \in \mathbb{R}^m the right-hand side , along with non-negativity \mathbf{x} \geq \mathbf{0}. This notation highlights the as an optimization over a defined by linear inequalities in . The corresponding dual problem is formulated as minimizing the dot product \mathbf{b} \cdot \mathbf{y}, where \mathbf{y} \in \mathbb{R}^m is the , subject to A^T \mathbf{y} \geq \mathbf{c} and \mathbf{y} \geq \mathbf{0}. Here, the transpose A^T ensures that the dual constraints mirror the primal's structure through the relationship, with the dual objective representing a weighted sum of resource values. This form underscores the between primal and dual, where the roles of objectives and constraints are interchanged via . Geometrically, the feasible set of the primal problem forms a in \mathbb{R}^n, defined as the of half-spaces corresponding to the inequalities A \mathbf{x} \leq \mathbf{b} and \mathbf{x} \geq \mathbf{0}. This polyhedral region is bounded or unbounded depending on the constraints, and optimal solutions occur at vertices where tight constraints intersect. Similarly, the feasible set is a in \mathbb{R}^m, providing a geometric on the primal's . The vector notation offers significant advantages for theoretical analysis, particularly in proving duality results. For instance, , which certifies the infeasibility of a system A \mathbf{x} \leq \mathbf{b}, \mathbf{x} \geq \mathbf{0} via a separating defined by a non-negative \boldsymbol{\lambda} such that A^T \boldsymbol{\lambda} \geq \mathbf{0} and \mathbf{b} \cdot \boldsymbol{\lambda} < 0, directly leverages this inner product structure to establish weak duality in linear programming. Such formulations enable elegant derivations of separation theorems and support extensions to convex optimization in spaces.

Matrix and Eigenvector Perspectives

The standard matrix formulation of a linear program (primal problem) is to maximize \mathbf{c}^T \mathbf{x} subject to \mathbf{A} \mathbf{x} = \mathbf{b} and \mathbf{x} \geq \mathbf{0}, where \mathbf{A} is an m \times n constraint matrix, \mathbf{b} \in \mathbb{R}^m is the right-hand side vector, \mathbf{c} \in \mathbb{R}^n is the objective coefficient vector, and \mathbf{x} \in \mathbb{R}^n is the decision variable vector. This equality-constrained form incorporates inequality constraints via slack variables, enabling a compact representation that highlights the linear structure and facilitates duality analysis. The corresponding dual problem minimizes \mathbf{b}^T \mathbf{y} subject to \mathbf{A}^T \mathbf{y} \geq \mathbf{c}, where \mathbf{y} \in \mathbb{R}^m are the dual variables (unrestricted in sign). From a spectral perspective, solutions to certain linear programs can be interpreted through generalized eigenvalue problems, particularly in the context of economic equilibrium models where duality arises naturally. In von Neumann's 1945 model of general economic equilibrium, optimal activity levels \mathbf{z} \geq \mathbf{0} and equilibrium prices \mathbf{p} > \mathbf{0} satisfy inequalities akin to dual constraints, with the maximal growth rate \rho (corresponding to the optimal value) serving as an eigenvalue. Specifically, the conditions \mathbf{p}^T \mathbf{A} \geq \rho \mathbf{p}^T and \mathbf{A} \mathbf{z} \leq \rho \mathbf{z} hold, with equality on the support of the optima, framing the solution as a pair of the matrix pencil (\mathbf{A}, \mathbf{I}). This viewpoint is insightful for theoretical analysis in specific models like von Neumann's, revealing connections between duality and eigenvalue problems in . However, it is not typically employed in computational algorithms for linear programs, as direct methods like the or interior-point approaches are more efficient for solving the matrix systems involved.

Examples

Basic Numerical Example

Consider the linear program in form: \begin{align*} \max &\quad 3x_1 + 4x_2 \\ \text{s.t.} &\quad x_1 + x_2 \leq 5, \\ &\quad 2x_1 + x_2 \leq 6, \\ &\quad x_1, x_2 \geq 0. \end{align*} The dual program is constructed by transposing the coefficient matrix of the constraints and switching the objective direction, yielding: \begin{align*} \min &\quad 5y_1 + 6y_2 \\ \text{s.t.} &\quad y_1 + 2y_2 \geq 3, \\ &\quad y_1 + y_2 \geq 4, \\ &\quad y_1, y_2 \geq 0. \end{align*} The feasible region for the primal is a polygon with vertices at (0,0), (0,5), (1,4), and (3,0). Evaluating the objective at these points gives values of 0, 20, 19, and 9, respectively, so the optimal primal solution is x_1 = 0, x_2 = 5, with objective value 20. For the dual, the feasible region has vertices at (4,0) and (0,4), with objective values of 20 and 24, respectively, so the optimal dual solution is y_1 = 4, y_2 = 0, with objective value 20. Weak duality holds, as for any feasible primal x and dual y, the primal objective satisfies $3x_1 + 4x_2 \leq 5y_1 + 6y_2. Strong duality is verified by the equal optimal values of 20. Complementary slackness conditions are satisfied: in the primal, the first constraint is tight (x_1 + x_2 = 5) corresponding to y_1 = 4 > 0, while the second has positive slack ($2x_1 + x_2 = 5 < 6) corresponding to y_2 = 0; in the dual, the second constraint is tight (y_1 + y_2 = 4) corresponding to x_2 = 5 > 0, while the first has positive slack (y_1 + 2y_2 = 4 > 3) corresponding to x_1 = 0. The following table summarizes the optimal solutions and related values:
AspectPrimal ValueDual Value
x_10-
x_25-
y_1-4
y_2-0
2020
Slack in 1st 01
Slack in 2nd 10

Resource Allocation Example

A classic illustration of duality in linear programming arises in resource allocation for agricultural production, such as a farmer deciding how many acres to allocate to corn (x_1) and (x_2) to maximize profit, subject to limits on and labor. The primal problem is formulated as maximizing profit from crop sales while respecting resource constraints: maximize $80x_1 + 60x_2, subject to x_1 + x_2 \leq 100 ( in acres), $2x_1 + x_2 \leq 150 (labor in hours), and x_1, x_2 \geq 0. The corresponding dual problem minimizes the imputed cost of the resources, where the dual variables y_1 and y_2 represent shadow prices for and labor, respectively: minimize $100y_1 + 150y_2, subject to y_1 + 2y_2 \geq 80 (for ), y_1 + y_2 \geq 60 (for ), and y_1, y_2 \geq 0. This dual formulation interprets the problem from the perspective of resource providers, seeking the minimum valuation of inputs that covers the profit contributions from each . Solving the yields an optimal solution of 50 acres of corn and 50 acres of , achieving a maximum of $7,000. The optimal solution is y_1 = 40 ( of $40 per acre of land) and y_2 = 20 ( of $20 per hour of labor), with a minimum cost of $7,000. By the duality , the maximum equals the minimum, confirming that the total value imputed to the exactly matches the farmer's at optimality. This economic interpretation highlights how dual shadow prices quantify the marginal value of scarce resources: an additional of would increase by $40, and an extra hour of labor by $20, guiding decisions on acquisition or trade-offs in crop planning.

Pathological Cases

In , pathological cases arise when the or problem lacks a finite optimum, either due to infeasibility or unboundedness. These situations highlight the reciprocal nature of duality: the and cannot both be feasible and bounded without attaining equal optimal values, but one may be infeasible while the other is unbounded. Such cases demonstrate the limitations of while preserving weaker properties. A classic example of an infeasible with an unbounded involves a simple one-variable maximization problem. Consider the problem: \begin{align*} \max &\quad x_1 \\ \text{s.t.} &\quad x_1 \geq 1, \\ &\quad x_1 \leq 0, \\ &\quad x_1 \geq 0. \end{align*} The constraints x_1 \geq 1 and x_1 \leq 0 (with x_1 \geq 0) form an empty , rendering the infeasible with no . Rewriting the explicit inequalities in standard form gives -x_1 \leq -1 and x_1 \leq 0. The corresponding , formulated in standard minimization form (noting that the non-negativity constraint does not introduce an additional variable), is: \begin{align*} \min &\quad -y_1 \\ \text{s.t.} &\quad -y_1 + y_2 \geq 1, \\ &\quad y_1, y_2 \geq 0. \end{align*} This dual is unbounded below; for instance, setting y_1 = t, y_2 = t + 1 for large t \geq 0 satisfies the constraint while the objective approaches -\infty as t \to \infty. Conversely, an unbounded primal pairs with an infeasible dual. Consider the primal: \begin{align*} \max &\quad x_1 \\ \text{s.t.} &\quad -x_1 \leq 0, \\ &\quad x_1 \geq 0. \end{align*} Here, the single constraint x_1 \geq 0 allows x_1 \to +\infty, making the primal unbounded above with objective value +\infty. The dual is: \begin{align*} \min &\quad 0 \cdot y_1 \\ \text{s.t.} &\quad -y_1 \geq 1, \\ &\quad y_1 \geq 0. \end{align*} The constraint -y_1 \geq 1 implies y_1 \leq -1, which contradicts y_1 \geq 0, so the dual is infeasible. In general, duality theory establishes this reciprocal behavior: if the is infeasible, the is either unbounded or infeasible, and if the is unbounded, the must be infeasible. The symmetric holds when swapping and roles. No finite optimum exists in these pairings, yet weak duality remains valid for any feasible points that may exist in one problem, providing bounds on the other's objective where applicable.

Algorithms and Computation

Primal-Dual Methods

Primal-dual interior-point methods represent a class of algorithms that solve problems by simultaneously optimizing both the and dual formulations, maintaining strict feasibility in the interior of the feasible regions. These methods trace a trajectory known as the central path, defined as the set of points (x_\tau, y_\tau, s_\tau) for \tau > 0 satisfying the primal constraints Ax = b, the dual constraints A^T y + s = c, and the centering condition X s = \tau e, where X = \operatorname{diag}(x) and e is the vector of ones. As \tau \to 0, these points approach an optimal solution satisfying complementary slackness. The core mechanism involves logarithmic barrier functions to enforce positivity, typically incorporating terms like -\sum \log x_i - \sum \log s_i, scaled by a barrier \mu = \tau. At each , the algorithm solves a perturbed Karush-Kuhn-Tucker (KKT) system using to compute search directions, where S = \operatorname{diag}(s): \begin{bmatrix} 0 & A & 0 \\ A^T & 0 & I \\ 0 & S & X \end{bmatrix} \begin{bmatrix} \Delta y \\ \Delta x \\ \Delta s \end{bmatrix} = \begin{bmatrix} r_p \\ r_d \\ r_c \end{bmatrix}, with centering residual r_c = \sigma \mu e - X s, where r_p, r_d are the and residuals, and \sigma \in [0,1] controls centering. A predictor-corrector variant, such as Mehrotra's method, first computes a pure centering direction (\sigma = 0) to predict the step, then adjusts \sigma adaptively for the corrector step to follow the central more closely. Under standard assumptions like strict feasibility and boundedness of the feasible sets, these methods exhibit polynomial-time , requiring O(\sqrt{n} \log(1/\epsilon)) iterations to achieve an \epsilon-optimal , where n is the number of variables. Long-step path-following variants reduce the duality measure \mu at a controlled rate, ensuring global to the central . Compared to pure primal interior-point methods, primal-dual approaches offer improved numerical conditioning by leveraging dual variables to balance the Newton systems, reducing sensitivity to ill-conditioned primal data and enhancing stability for large-scale problems. This duality integration also allows early detection of primal or dual infeasibility through residual monitoring.

Dual Simplex Algorithm

The dual simplex algorithm is a variant of the simplex method for solving linear programming problems, introduced by C. E. Lemke in 1954 as a computational procedure that mirrors the simplex but operates on the dual formulation. Unlike the standard simplex method, which begins with a primal feasible but dual infeasible basis and proceeds toward dual feasibility, the dual simplex starts from a dual feasible but primal infeasible basis and iterates toward primal feasibility while preserving dual feasibility. This approach ensures that the reduced costs remain non-negative (for a maximization primal), maintaining optimality conditions for the dual problem throughout the process. The procedure begins with a basic solution where the dual variables y = B^{-T} c_B satisfy the dual constraints (dual feasibility), but the primal basic solution \bar{b} = B^{-1} b may have negative entries, indicating primal infeasibility. To proceed, select a leaving variable: choose the row index i corresponding to the most negative entry in \bar{b} (or any negative \bar{b}_i < 0). For the entering variable, consider non-basic columns j where the entry \bar{a}_{ij} < 0 (to ensure the pivot improves feasibility); compute the ratios \frac{\bar{c}_j}{|\bar{a}_{ij}|} (where \bar{c}_j \geq 0 maintains dual feasibility), and select the j with the minimum ratio to enter the basis. Perform the pivot by updating the basis matrix B to \hat{B} = B - \{i\} + \{j\}, transforming the tableau via on the pivot element \bar{a}_{ij}. Repeat these steps until \bar{b} \geq 0, at which point the solution is primal and dual feasible, hence optimal by the duality theorem. If at any iteration no suitable entering variable exists (i.e., \bar{a}_{ij} \geq 0 for all j when \bar{b}_i < 0), the problem is infeasible. The algorithm can be implemented in revised form, storing only the basis inverse B^{-1} and updating it via rank-one modifications, which enhances efficiency for large-scale problems similar to the revised . Key use cases include post-optimality analysis, such as when right-hand side values change (e.g., adding resources or constraints), rendering the current solution infeasible but leaving the feasible due to unchanged objective coefficients. This makes it particularly valuable for in applications like , where incremental modifications are common. The method terminates at optimality through : once feasibility is achieved alongside maintained feasibility, complementary slackness holds, confirming the solution is both and optimal.

Applications and Implications

Classical Theorems and Optimizations

One of the foundational applications of duality lies in establishing min-max equalities for network problems. The asserts that in a capacitated , the maximum value of a from a to a equals the minimum of a cut separating from the . This result can be derived by formulating the maximum as a linear , where variables represent along edges subject to and constraints, and the minimum cut as its , with dual variables associated with nodes indicating assignments. ensures the optimal values coincide, providing a rigorous proof of the equality. This duality perspective not only proves the theorem but also highlights the structural interplay between flows and cuts, influencing subsequent developments in . The original formulation and for computing maximum flows, which underpin this theorem, were introduced by and Fulkerson in , who demonstrated the equivalence through augmenting methods that align with dual optimality conditions. In , duality similarly underpins König's for bipartite graphs, which states that the size of the maximum matching equals the size of the minimum . The maximum matching problem is modeled as a primal linear program with variables indicating edge selections, constrained by degree limits at vertices, while the minimum serves as the , with variables representing vertex coverage costs. The integrality of the bipartite matching polytope ensures that optimal solutions are integral, and duality yields the equality without requiring separate combinatorial arguments. König's original combinatorial proof from 1931 relied on alternating paths, but the linear programming duality approach, popularized in the mid-20th century, offers a unified algebraic verification that extends to related covering problems. This theorem exemplifies how dual programs provide lower bounds that match primal optima in structured settings, such as problems. Von Neumann's extends duality to , proving that for any finite zero-sum two-player game with payoff matrix A, the value v satisfies \max_x \min_y x^T A y = \min_y \max_x x^T A y = v, where x and y are mixed strategies. This is formulated as a primal-dual pair of linear programs: the primal maximizes the expected payoff for the row player subject to strategy normalization, and the dual minimizes it for the column player. The theorem guarantees the existence of optimal strategies and equal values, with duality providing the equivalence. Proved by in 1928 using fixed-point arguments, the result was later reinterpreted through in the 1950s, revealing that solving zero-sum games reduces to linear programming feasibility and optimization. This connection has profound implications for computation in strategic settings. Hoffman's circulation theorem addresses the existence of feasible circulations in networks with lower and upper bounds on edge flows, stating that a circulation exists if and only if for every cycle, the sum of lower bounds does not exceed the sum of upper bounds adjusted by the cycle direction. This is established via duality: the primal seeks a circulation satisfying bound constraints and conservation at nodes, while the dual involves potentials and cycle multipliers that certify infeasibility through Farkas-like alternatives. Duality ensures that if the primal is infeasible, the dual provides a separating witnessing the obstruction. Hoffman introduced this theorem in 1960, leveraging linear inequalities to generalize earlier flow existence results, and it remains a cornerstone for bounded network problems in .

Modern Uses in Machine Learning

In support vector machines (SVMs), the formulation plays a central role in enabling methods for non-linear classification. The problem minimizes \frac{1}{2} \|w\|^2 + C \sum_{i=1}^n \xi_i subject to y_i (w \cdot x_i + b) \geq 1 - \xi_i and \xi_i \geq 0, where C controls the trade-off between margin maximization and classification error. The corresponding is \max_{\alpha} \sum_{i=1}^n \alpha_i - \frac{1}{2} \sum_{i,j=1}^n \alpha_i \alpha_j y_i y_j \langle x_i, x_j \rangle subject to \sum_{i=1}^n \alpha_i y_i = 0 and $0 \leq \alpha_i \leq C, which depends only on inner products \langle x_i, x_j \rangle. Replacing these with a K(x_i, x_j) allows implicit mapping to high-dimensional feature spaces without explicit computation, facilitating scalable non-linear SVMs widely used in tasks like image recognition and bioinformatics. The of regression, an \ell_1- method for sparse , provides insights into variable importance and efficient computation. The primal problem is \min_{\beta} \frac{1}{2} \|y - X\beta\|_2^2 + \lambda \|\beta\|_1, and its is \max_{v} -\frac{1}{2} \|v\|_2^2 + y^T v subject to \|X^T v\|_\infty \leq \lambda. The variables v represent residuals adjusted for the regularization , and the condition \|X^T v\|_\infty \leq \lambda enforces sparsity by zeroing coefficients where the exceeds the . This duality enables path-following algorithms for \lambda and has been applied in high-dimensional and to identify key predictors. In , duality principles from extend to quadratic programs like the Markowitz mean-variance model for analyzing risk-return trade-offs in applications informed by . The primal quadratic program minimizes w^T \Sigma w subject to \mu^T w \geq r, \sum w = 1, and w \geq 0, where w are asset weights, \Sigma is the , and \mu are expected returns. The Lagrangian dual introduces multipliers \lambda for the return constraint and \nu for the , yielding \min_{\lambda \geq 0, \nu} \lambda r + \nu subject to \Sigma^{-1} (\lambda \mu - \nu \mathbf{1}) \geq 0, which interprets \lambda as the shadow price of return and aids in robust estimation under uncertainty. This framework integrates with techniques like for dynamic allocation, improving out-of-sample performance in . Recent applications of dual formulations in include adversarial and generative adversarial networks (GANs). In adversarial for SVMs, Lagrange duality certifies robustness by solving a semi-infinite program that bounds perturbations within an \ell_p-ball, reformulating the as a tractable dual problem to enhance model resilience against attacks in cybersecurity and autonomous systems. For GANs, convex duality frameworks analyze the objective \min_G \max_D V(D,G) = \mathbb{E}_{x \sim p_r} [\log D(x)] + \mathbb{E}_{z \sim p_z} [\log (1 - D(G(z)))], providing tight lower bounds on divergences between real and generated distributions via Fenchel conjugates, which guide stable and evaluate convergence in image synthesis tasks.

References

  1. [1]
    [PDF] Duality in Linear Programming - MIT
    Duality in linear programming is essentially a unifying theory that develops the relationships between a given linear program and another related linear ...
  2. [2]
    [PDF] Lecture 6 1 The Dual of Linear Program - Stanford CS Theory
    Jan 20, 2011 · In which we introduce the theory of duality in linear programming. 1 The Dual of Linear Program. Suppose that we have the following linear ...
  3. [3]
    [PDF] DRAFT Formulation and Analysis of Linear Programs
    Sep 5, 2016 · programming. A prime application of linear programming is to the allocation of limited resources among production activities that can be ...
  4. [4]
    [PDF] Linear programming 1 Basics
    Mar 17, 2015 · Linear Programming deals with the problem of optimizing a linear objective function subject to linear equality and inequality constraints on ...
  5. [5]
    None
    Below is a merged and comprehensive summary of the geometric interpretation of the dual problem in linear programming (Section 5.3) based on all provided segments. To retain all information in a dense and organized manner, I will use a combination of narrative text and a table in CSV format to capture key details, equations, and concepts efficiently. The narrative will provide an overarching explanation, while the table will summarize specific details such as concepts, equations, and references.
  6. [6]
    [PDF] Computing True Shadow Prices in Linear Programming. - DTIC
    It is well known that in linear programming, the optimal values of the dual variables can be interpreted as shadow prices (marginal values) of the.Missing: seminal | Show results with:seminal
  7. [7]
    [PDF] 4 Duality Theory
    That is, the dual problem is unbounded below, and so, by the weak duality theorem, the primal problem must be infeasible! We will make extensive use of the dual ...Missing: definition | Show results with:definition
  8. [8]
    [PDF] 1 Outline 2 Deriving the dual linear program - Dabeen Lee
    Mar 28, 2023 · Let us outline the procedure of deriving the dual LP with the above general LP. Step 1: assign dual variables to each constraint, except sign ...
  9. [9]
    [PDF] Lecture 10: Duality in Linear Programs 10.1 Lower Bounds in Linear ...
    Generally, the number of dual variables is equal to the number of inequality constraints and equality con- straints in the primal problem.
  10. [10]
    [PDF] 0.1 Weak LP Duality - cs.Princeton
    If optimum of LP1 finite, then the optimum of LP2 is also finite, and they are equal. The key ingredient in the proof will be what's called the Separating ...
  11. [11]
    [PDF] Duality 1.1 Structure of LP solutions
    This lecture covers weak and strong duality, and also explains the rules for finding the dual of a linear program, with an example. Before we move on to duality ...<|control11|><|separator|>
  12. [12]
    [PDF] Lecture 8 1 Strong duality
    Sep 18, 2014 · Let x and y be feasible for the primal and dual, respectively. Recall the proof of weak duality: cT x = n. X j=1 cjxj ≥ n. X j=1 m. X i=1.
  13. [13]
    [PDF] ‣ LP duality ‣ strong duality theorem ‣ bonus proof ... - cs.Princeton
    Jul 25, 2017 · LINEAR PROGRAMMING II. ‣ LP duality. ‣ strong duality theorem ... ‣ bonus proof of LP duality. ‣ applications. Page 24. Strong duality theorem.
  14. [14]
    [PDF] chapter xix - linear programming and the theory of games
    *This chapter was presented in a preliminary form by A. W. Tucker at a meeting of the Econometric Society at Boulder, Colorado, September 2, 1949. * Under ...Missing: paper | Show results with:paper
  15. [15]
    [PDF] 1 LP Geometry
    LP geometry involves hyperplanes, convex polyhedra, and vertices. A vertex is a point that is an endpoint of any line segment within the polyhedron.
  16. [16]
    [PDF] 1 Linear Programming and Farkas' Lemma - cs.Princeton
    Then the Linear Program (LP for short) can be rewritten as: min cT X : AX ≥ b. X ≥ 0. (3). This form is general enough to represent any possible linear program.
  17. [17]
    [PDF] Chapter 1: Linear Programming
    May 22, 2013 · Similar other constraints, Farmer's perspective yields original MLP. Chapter 1: Linear Programming. 44. Page 45. Duality: Introductory Example, ...
  18. [18]
    [PDF] Linear Programming Duality Theorem from the ... - Lehigh University
    Weak duality shows that if the primal or dual is unbounded then the other must be infeasible. Thus there are four possibilities for a primal-dual pair: both ...
  19. [19]
    [PDF] Duality Theory
    Dual Theorem: An LP has an optimal solution if and only if its dual has an optimal solution, and in this case their optimal values are equal. An immediate ...
  20. [20]
    Primal-Dual Interior-Point Methods | SIAM Publications Library
    This book presents the major primal-dual algorithms for linear programming in straightforward terms. A thorough description of the theoretical properties of ...Missing: seminal | Show results with:seminal
  21. [21]
    On the Implementation of a Primal-Dual Interior Point Method
    This paper gives an approach to implementing a second-order primal-dual interior point method. It uses a Taylor polynomial of second order to approximate a ...
  22. [22]
    [PDF] Linear Programming: Interior-Point Methods - cs.wisc.edu
    By the early 1990s, one class—primal-dual methods— had distinguished itself as the most efficient practical approach, and proved to be a strong competitor to ...
  23. [23]
    The dual method of solving the linear programming problem - Lemke
    This paper develops a new computational method of solving general linear programming problems. References 1 Charnes, A., and Lemke, CE, “ A Modified Simplex ...Missing: algorithm | Show results with:algorithm
  24. [24]
    [PDF] Lecture 15 1 Varieties of Simplex Method: Dual Simplex
    Dual simplex is exactly analogous to primal simplex where we start with a dual feasible solution corresponding to a basis B and move towards making the ...
  25. [25]
    The Dual Simplex Method
    This new pivoting strategy is called the Dual Simplex Method because it really is the same as performing the usual Simplex Method on the dual linear problem.
  26. [26]
    The Dual Simplex Algorithm | SpringerLink
    A procedure that solves a dual linear programming problem may be called a dual simplex algorithm. This algorithm was discovered by CE Lemke in 1954.
  27. [27]
    [PDF] 1. Lecture notes on bipartite matching
    Feb 8, 2011 · Theorem 1.1 (König 1931) For any bipartite graph, the maximum size of a matching is equal to the minimum size of a vertex cover. We shall prove ...Missing: source | Show results with:source
  28. [28]
  29. [29]
    Support-vector networks
    The support-vector network is a new leaming machine for two-group classification problems. The machine conceptually implements the following idea: input vectors ...
  30. [30]
    [PDF] On the LASSO and Its Dual
    Thus, the LASSO combines characteristics of ridge regression and subset selection and promises to be a useful tool for variable selection. Though the ...
  31. [31]
    [PDF] Lecture 5: Mean variance portfolio optimization & Lagrange Duality
    Problem (1) is called the basic version of the Markovitz portfolio optimization problem. We present two properties of the covariance matrix. 1. Page 2 ...
  32. [32]
    [PDF] Adversarial Robustness with Semi-Infinite Constrained Learning
    In particular, we leverage semi-infinite optimization and non-convex duality theory to show that adversarial training is equivalent to a statistical problem ...<|separator|>
  33. [33]
  34. [34]
    [PDF] A Convex Duality Framework for GANs - NIPS papers
    We also discuss the application of our duality framework to neural net discriminators with bounded Lipschitz constants. While a set of neural network functions ...Missing: bounds | Show results with:bounds