Dual of a linear program

Dependencies:

1. Optimization: Dual and Lagrangian

Let $A \in \mathbb{R}^{m \times n}$. Let $I \subseteq [m]$ and $J \subseteq [n]$. Then the dual of the linear program \[ P: \min_{x \in \mathbb{R}^n: x_J \ge 0} c^Tx \quad\textrm{where}\quad ((Ax)_i \ge b_i, \forall i \in I) \textrm{ and } ((Ax)_i = b_i, \forall i \in [m]-I) \] is \[ D: \max_{y \in \mathbb{R}^m: y_I \ge 0} b^Ty \quad\textrm{where}\quad ((A^Ty)_j \le c_j, \forall j \in J) \textrm{ and } ((A^Ty)_j = c_j, \forall j \in [n]-J) \] Also, the dual of $D$ is $P$.

As a special case, on setting $J = [n]$ and $I = [m]$, we get that the dual of \[ P: \min_{x \ge 0} c^Tx \textrm{ where } Ax \ge b \] is \[ D: \max_{y \ge 0} b^Ty \textrm{ where } A^Ty \le c \]

Equivalently, we can obtain the dual by applying this method: http://www.cs.columbia.edu/coms6998-3/lpprimer.pdf.

Dependency for:

1. Linear programming: weak duality
2. Linear programming: strong duality (incomplete)
3. AnyPrice share
4. Bin packing: dual of config LP
5. Bin packing: dual of the density-restricted config LP

Info:

• Depth: 1
• Number of transitive dependencies: 1

Transitive dependencies:

1. Optimization: Dual and Lagrangian