Weak Duality Theorem

Statement and consequences of the Weak Duality Theorem for linear programs: every dual feasible objective bounds every primal feasible objective, yielding bounds on optima, infeasibility/unboundedness corollaries, and a sufficient optimality certificate.

Step 1 of 157%

Tutorial

The Weak Duality Theorem

Recall the standard-form primal LP and its dual:

(P):max cxs.t.Axb, x0.\text{(P):}\quad \max\ \mathbf{c}^\top\mathbf{x}\quad \text{s.t.}\quad A\mathbf{x}\le \mathbf{b},\ \mathbf{x}\ge \mathbf{0}. (D):min bys.t.Ayc, y0.\text{(D):}\quad \min\ \mathbf{b}^\top\mathbf{y}\quad \text{s.t.}\quad A^\top\mathbf{y}\ge \mathbf{c},\ \mathbf{y}\ge \mathbf{0}.

The Weak Duality Theorem states that for every primal feasible x\mathbf{x} and every dual feasible y\mathbf{y},

cx  by.\mathbf{c}^\top\mathbf{x}\ \le\ \mathbf{b}^\top\mathbf{y}.

Proof. Since x0\mathbf{x}\ge\mathbf{0} and AycA^\top\mathbf{y}\ge \mathbf{c}, we have cx(Ay)x=y(Ax)\mathbf{c}^\top\mathbf{x}\le (A^\top\mathbf{y})^\top\mathbf{x}=\mathbf{y}^\top(A\mathbf{x}). Since y0\mathbf{y}\ge\mathbf{0} and AxbA\mathbf{x}\le\mathbf{b}, we have y(Ax)yb=by\mathbf{y}^\top(A\mathbf{x})\le \mathbf{y}^\top\mathbf{b}=\mathbf{b}^\top\mathbf{y}. Chaining the two gives the result. \blacksquare

Tiny illustration. Let c=[32]\mathbf{c}=\begin{bmatrix}3\\2\end{bmatrix}, b=[45]\mathbf{b}=\begin{bmatrix}4\\5\end{bmatrix}, A=[1112]A=\begin{bmatrix}1&1\\1&2\end{bmatrix}. Take x=[11]\mathbf{x}=\begin{bmatrix}1\\1\end{bmatrix} (primal feasible: Ax=[23]bA\mathbf{x}=\begin{bmatrix}2\\3\end{bmatrix}\le\mathbf{b}) and y=[21]\mathbf{y}=\begin{bmatrix}2\\1\end{bmatrix} (dual feasible: Ay=[34]cA^\top\mathbf{y}=\begin{bmatrix}3\\4\end{bmatrix}\ge\mathbf{c}). Then

cx=3+2=5,by=8+5=13,\mathbf{c}^\top\mathbf{x}=3+2=5,\qquad \mathbf{b}^\top\mathbf{y}=8+5=13,

and indeed 5135\le 13.

navigate · Enter open · Esc close · ⌘K/Ctrl K toggle