I have the following optimization problem:

\begin{alignat}2\min &\quad A(t)\cdot x(t)-B(t)\cdot y(t)+C(t)\cdot z(t)-D(t)\cdot k(t)\\\text{s.t.}&\quad z(t)+z_1(t)-y(t)-y_1(t)+x(t) = k(t);& \lambda(t)\\&\quad0 \le x(t) \le \bar{X};& \mu^1(t),\mu^2(t)\\&\quad0 \le k(t) \le \bar{K};& \mu^3(t),\mu^4(t)\\&\quad0 \le y(t) \le \bar{Y};& \mu^5(t),\mu^6(t)\\&\quad0 \le z(t) \le \bar{Z};& \mu^7(t),\mu^8(t)\\&\quad0 \le e(t) \le \bar{E};& \mu^9(t),\mu^{10}(t)\\&\quad e(t) = e(t-1)-(z+z_1)+\beta(y+y_1);&\mu^{11}(t)\\&\quad e(1) = 0;&\mu^{12}(t)\end{alignat}

The dual variables are listed against each equation. All the capital letters and letters under the bar are parameters and the rest are variables. Then, I derive the KKT optimality conditions for this problem by taking the Lagrangian. When I take the derivative of the Lagrangian w.r.t $z_1$ and $y_1$, I get the following primal feasibility conditions.

\begin{alignat}2-\lambda(t)+\mu^{11}(t) &= 0 \tag1\\\lambda(t)-\beta\cdot\mu^{11}(t) &= 0 \tag2\end{alignat}

where, $\beta = 0.85$. My question is - what does it mean to have one dual variable ($\lambda(t)$ in this case) equating to two different values as shown in $(1)$ and $(2)$?

  • 1
    $\begingroup$ Are $z_1(t)$ and $y_1(t)$ unbounded free variables? $\endgroup$
    – prubin
    Commented Nov 25, 2019 at 19:07
  • $\begingroup$ Sorry, I forgot to include, they are non-negative variables. $\endgroup$
    – S_Scouse
    Commented Nov 25, 2019 at 19:19
  • $\begingroup$ I interpret what you have here to mean that the optimal point has mu11 == 0. $\endgroup$
    – Brannon
    Commented Oct 27, 2021 at 16:59


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.