Content deleted Content added
→Relation to quadratic programming: added clarification on the conversion QP -> LCP |
→Relation to quadratic programming: further clarification n the conversion from QP to LCP |
||
Line 37:
* <math>\mathbf{q} = \left[\begin{array}{c}\mathbf{c}\\-\mathbf{b}\end{array}\right]</math>
* <math>\mathbf{M} = \left[\begin{array}{cc} \mathbf{Q} & -\mathbf{A}^{T}\\ \mathbf{A} & 0\end{array}\right]</math>
This is because the [[Karush-Kuhn-Tucker]] conditions of the QP problem can be written as:
* <math>\mathbf{v} = \mathbf{Q} \mathbf{x} - \mathbf{A}^{T} \mathbf{mu} + \mathbf{c}</math>
* <math>\mathbf{s} = \mathbf{A} \mathbf{x} - \mathbf{b}</math>
* <math>\mathbf{x}, \mathbf{\mu}, \mathbf{v}, \mathbf{s} \ge \mathbf{0}</math>
* <math>\mathbf{x}^{T}\mathbf{v} + \mathbf{\mu}^{T}\mathbf{s} = \mathbf{0}</math>
...being <math> \mathbf{
In that case,
Line 42 ⟶ 55:
* <math>\mathbf{z} = \left[\begin{array}{c}\mathbf{x}\\ \mathbf{\mu}\end{array}\right]</math>
* <math>\mathbf{w} = \left[\begin{array}{c}\mathbf{v}\\ \mathbf{s}\end{array}\right]</math>
▲...being <math>\mathbf{\mu}</math> the multipliers on the inequality constraints, <math>\mathbf{v}</math> the multipliers on the non-negativity constraints, and <math> \mathbf{s} {=} \mathbf{A}\mathbf{x} - \mathbf{b} </math> the slack variables for the inequality constraints.
In fact, most QP solvers work on the LCP formulation, including the [[interior point method]], principal / complementarity pivoting and [[active set]] methods.
|