\begin{equation}\label \arg\min\limits_{\mathbf{c}} \mathbf{c}^TA\mathbf{c}-\mathbf{b}^T\mathbf{c}\\ s.t. \quad c_i = k_i, \quad i\in I \end{equation}

where $A\in R^{M\times M}$ is a large sparse semi-positive matrix, $\mathbf{c} \in R^{M \times 1}$, and $k_i$ are given constants, $I\subseteq \{1,2,\ldots,M\}$.

How to solve this quadratic constrained problem using MATLAB or CVX software effectively?

  • $\begingroup$ Matlab has QuadProg. If the problem is really large you may want to consider a high-performance commercial solver such as Cplex, Gurobi or Mosek. I think they all have Matlab bindings of some sort. $\endgroup$ – Erwin Kalvelagen Oct 4 '16 at 10:29

If you simply fix the variables which you now have written as equality constraints, you can reduce the objective in the remaining free variables $\tilde{c}$ to $\tilde{c}^T\tilde{A}\tilde{c} + \tilde{b}^T\tilde{c} + f$, and the optimal solution is given by the solution to the stationarity condition $2\tilde{A}\tilde{c} + \tilde{b} = 0$, i.e., a linear system of equations. No reason to use a general solver here.

  • $\begingroup$ The $\tilde{c}$ is missing in your stationarity condition: $2\tilde{A}\tilde{c}+\tilde{b} = 0$ $\endgroup$ – LinAlg Oct 4 '16 at 14:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.