1
$\begingroup$

I'm studying dimensionality reduction (SVD in particular), and I saw the following question:

Assume we have a vector $x \in \mathbb R^d$, and consider $F(x)=s^t x$ , where $s$ is a $d$-dimensional random vector with entries drawn uniformly independently from $[-1,1]$.

What is the value of $\mathbb E[F(x)^2]$?

I'm starting now from zero, so I need to study more. The exercise asks for a formal proof, but I wish to just understand the philosophy.

I see many questions in which valued are drawn uniformly and independently from $[-1,1]$. Is this distribution used due to its symmetrical range? Is the expectation in this range often $1/2$?

$\endgroup$
1
$\begingroup$

Let us assume that each entry of $s$ is drawn independently from some distribution $\mathcal{D}$ whose expectation is $0$. Then $$ \mathbb{E}\left[\left(\sum_{i=1}^d s_i x_i\right)^2\right] = \sum_{i=1}^d x_i^2 \mathbb{E}[s_i^2] + \sum_{i \neq j} x_i x_j \mathbb{E}[s_i s_j] = \|x\|^2 \mathbb{V}[\mathcal{D}], $$ since $\mathbb{E}[s_is_j] = \mathbb{E}[s_i]\mathbb{E}[s_j] = 0$ due to independence.

In your particular case, $\mathcal{D}$ is the uniform distribution over $[-1,1]$, whose variance is $$ \frac{1}{2} \int_{-1}^1 x^2 \, \mathrm{d}x = \left. \frac{x^3}{6} \right|_{-1}^1 = \frac{1}{3}. $$

$\endgroup$
  • $\begingroup$ Expectaction of $ F(X) $is 0 ...right? instead of $[F(x)]^2$ is the $|| v ||^2$ . $\endgroup$ – theantomc Jan 8 at 11:42
  • $\begingroup$ The expectation of $F(x)$ is zero. I didn’t understand the rest of your question. $\endgroup$ – Yuval Filmus Jan 8 at 11:45

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.