Let $\vec{a}=(a_1,a_2,\dots,a_n)$ be a given vector that satisfies $a_1+a_2+...+a_n=0$. Let $\vec{x}=(x_1,x_2,\dots,x_n)$ be an arbitrary vector that satisfies $x_1+x_2+...+x_n=0$, but we know $\|\vec{x}\|_2$.

Is there a tight upper bound for $(\vec{a}^T \cdot \vec{x})$ in terms of $a_1, a_2, \cdots, a_n$, and $ \|\vec{x}\|_2$ ?

I know Cauchy-Schwarz inequality can produce an upper bound:

$$|\vec{a}^T \cdot \vec{x} | \le \|\vec{a}\|_2 \|\vec{x}\|_2 .$$

However, this bound seems not very tight and does not consider my constraints that $x_1+x_2+\dots+x_n=0$ and $a_1+a_2+\dots+a_n=0$. Any suggestion is appreciated.

  • 1
    $\begingroup$ What do you mean that the Cauchy-Schwarz bound is not tight? It is saturated when x is a multiple of a, which seems to be allowed by your setup. $\endgroup$
    – Mike Hawk
    Mar 13 '18 at 20:24

The bound may not seem tight, but take $$\vec{y} = \frac{\vec{a}}{\|\vec{a}\|_2} \|\vec{x}\|_2$$ Observe $\vec{y}$ satisfies both your conditions, and $|\vec{a}^T \cdot \vec{y}| = \|\vec{a}\|_2\|\vec{x}\|_2$, so the bound is tight.


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.