In the context of Lagrangian relaxation of discrete optimization problems, what does it mean to 'dualize a constraint'?
Asked
Active
Viewed 1,177 times
1 Answers
0
You have a function $f(\vec x)$ you wish to optimise, given some constraints $g_i(\vec x)=c_i$. Then you can write a Lagrangian $$L=f-\sum_i \lambda_i(g_i-c_i)$$Then to dualise this means to rewrite it as a problem where you optimise a function $F(\vec\lambda)=\sum_i\lambda_ic_i$, with respect to some constraints $G_i(\vec\lambda)=C_i$. When rewriting the problem like this, the components $x_i$ become the Lagrange multipliers for the Lagrangian for this dual problem.
John Doe
- 14,765