Can ADMM be applied to optimisation with matrix variables and inequality constraint?

278 Views Asked by At

Can ADMM be applied to optimisation with matrix variables and inequality constraint? If so, how may I do it? My optimisation problem looks something like this:

\begin{align} \min_{X \in \mathbb{R}^{n \times n}} \quad & \|Xv \|_2 + \|X\|_F^2 \\ \quad &\text{Tr}(AX) \geq \text{Tr}(B) \end{align} where $v \in \mathbb{R}^n$, $A,B \in \mathbb{R}^{n \times n}$ are known constant vector and matrices.

Thanks a lot!

1

There are 1 best solutions below

0
On

Honestly, I do not see big issues here. Assume $n=2$ to clarify the ideas, and set $$ v = \begin{pmatrix} v_1 \\ v_2 \end{pmatrix}, \ A= \begin{pmatrix} a_{11} & a_{12} \\ a_{21} & a_{22} \end{pmatrix}, \ B= \begin{pmatrix} b_{11} & b_{12} \\ b_{21} & b_{22} \end{pmatrix}, \ X= \begin{pmatrix} x_{11} & x_{12} \\ x_{21} & x_{22} \end{pmatrix}. $$Then $$ \|Xv\|_2 + \|X\|_F ^2 = \sqrt{ (v_1 x_{11} + v_2 x_{12} )^2 + (v_1 x_{21} + v_2 x_{22})^2 }+ x_{11} ^2 + x_{12} ^2 + x_{21} ^2 + x_{22} ^2, $$ $$\text{Tr}(AX) = a_{11} x_{11} + a_{12} x_{21} + x_{22} a_{22} + a_{21} x_{12} $$ and $$\text{Tr}(B)=b_{11} + b_{22}. $$

Then you want to minimize a nonlinear objective function (differentiable almost everywhere, but maybe you wanted $\|Xv\|_2 ^2$ instead?) under linear inequality constraints. This paper briefly explains how ADMM can deal with linear inequality constraints.