Prove minimum for constrained convex optimization.

27 Views Asked by At

I am given the following:

min $f(x) = 1/2(x-r)^T(x-r)$

s.t $a^Tx=b$

This is basically minimizing the distance from x on the hyperplan $a^Tx = b$ from some point r in $R^n$

If I compute the gradient I get:

$\nabla f(x) = x - r$ and set = $a\lambda \implies x = r + a\lambda$

However the book states the minim value $x_*$ is given by:

$x_* = r + \frac{b-a^Tr}{a^Ta}a$

I'm not sure how they're getting this or how to show $\lambda = \frac{b-a^Tr}{a^Ta}$

1

There are 1 best solutions below

1
On BEST ANSWER

$$x-r=a\lambda$$

Multiplying $a^T$ on both sides,

$$a^T(x-r)=\lambda (a^Ta)$$

$$a^Tx-a^Tr=\lambda (a^Ta)$$

But we know that $a^Tx=b$.

$$b-a^Tr=\lambda (a^Ta)$$