Linearly Constrained Least-Squares Problems

Least-Squares > Definitions >> [ LS | Regularized LS | Constrained LS ] | Solution | Sensitivity | Limitations
  • Linearly-constrained least-squares
  • Minimum-norm solutions to linear equations

Linearly constrained least-squares

Definition

An interesting variant of the ordinary least-squares problem involves equality constraints on the decision variable x:

\min\limits_x ||Ax-y||_2^2:Cx=d,

where C \in R^{p \times n}, and d \in R^p are given.

Examples:

Solution

We can express the solution by first computing the nullspace of C. Assuming that the feasible set of the constrained LS problem is not empty, it can be expressed as

X={x_0+Nz : z \in R^k},

where k is the dimension of the nullspace of C, and x_0 is a particular solution to the equation Cx=d.

Expressing x in terms of the free variable z, we can write the constrained problem as an unconstrained one:

\min\limits_{x} ||\tilde{A}z-\tilde{y}||_2

where \tilde{A}:=AN, and \tilde{y}=y-Ax_0.

Minimum-norm solution to linear equations

A special case of linearly constrained LS is

\min\limits_x ||x||_2 : Ax = y,

in which we implicitly assume that the linear equation in x: Ax=y, has a solution, that is, y is in the range of A.

The above problem allows selecting a particular solution to a linear equation, in the case when there are possibly many, that is, the linear system aX=Y is under-determined.

As seen here, when A is full row rank, the matrix AA^T is invertible, and the above has the closed-form solution

x^*=A^T(AA^T)^{-1}y.

ExamplesControl positioning of a mass.

License

Hyper-Textbook: Optimization Models and Applications Copyright © by L. El Ghaoui. All Rights Reserved.

Share This Book