A Are these two optimization problems equivalent?

Click For Summary
The discussion revolves around the equivalence of two optimization problems, P1 and P2, involving complex variables and constraints. The first problem aims to maximize the squared difference between a complex scalar d and a sum involving variables z_n, subject to a specific constraint. The second problem reformulates the first by introducing new variables y_n, making it potentially easier to solve using semidefinite programming. The main inquiry is whether these two problems are equivalent in terms of their solutions and constraints. The thread concludes with a request to delete the post due to an early submission error.
haji-tos
Messages
5
Reaction score
1
Hello,

I need help please. I have the following optimization problem defined as

\begin{equation}
\begin{aligned}
& (\mathbf{P1}) \quad \max_{\mathbf{z}} \quad \left| d -\sum_{n=1}^{N} \frac{c_n}{f_n + z_n} \right|^2 \\
& \text{subject to} \quad \sum_{n=1}^{N} \frac{|a_n|^2 \text{Re}(z_n)}{|f_n + z_n|^2} = 0.
\end{aligned}
\end{equation}
where d is a complex scalar, f=[f1,...,fN], c=[c1,...,cN] and a=[a1,...,aN] are complex vectors.

I am trying to solve this so I was thinking to consider

\begin{equation}
y_n=\frac{1}{f_n+z_n}, \quad \forall n \in \{1,...,N\}
\end{equation}
and
\begin{equation}
z_n=\frac{1}{y_n}-f_n, \quad \forall n \in \{1,...,N\}
\end{equation}

and then transform the problem into

\begin{equation}
\begin{aligned}
& (\mathbf{P2}) \quad \max_{\mathbf{y}} \quad \left| d -\sum_{n=1}^{N} c_n y_n \right|^2 \\
& \text{subject to} \quad \sum_{n=1}^{N} |a_n|^2 \text{Re}(y_n^* - f_n y_n y_n^*) = 0.
\end{aligned}
\end{equation}
which is easier to solve using semidefinite programming.
Can you please tell me if the two problems are equivalent ?

Thank you very much !
 
Mathematics news on Phys.org
Please delete this post... I made a mistake of posting too early, thinking the LATEX was not showing up. Thanks.
 
Thread 'Erroneously  finding discrepancy in transpose rule'
Obviously, there is something elementary I am missing here. To form the transpose of a matrix, one exchanges rows and columns, so the transpose of a scalar, considered as (or isomorphic to) a one-entry matrix, should stay the same, including if the scalar is a complex number. On the other hand, in the isomorphism between the complex plane and the real plane, a complex number a+bi corresponds to a matrix in the real plane; taking the transpose we get which then corresponds to a-bi...

Similar threads

  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 2 ·
Replies
2
Views
1K
  • · Replies 3 ·
Replies
3
Views
2K
  • · Replies 8 ·
Replies
8
Views
2K
  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 6 ·
Replies
6
Views
2K
  • · Replies 18 ·
Replies
18
Views
3K
Replies
6
Views
2K
  • · Replies 21 ·
Replies
21
Views
1K
  • · Replies 8 ·
Replies
8
Views
1K