3
$\begingroup$

Let $\mathbb{F}$ be the field of size 2. For a function $f : \mathbb{F}^n \to \mathbb{F}$, let $d(f)$ be the smallest integer such that there exists a degree-$d(f)$, $n$-variate, multilinear polynomial that induces $f$ on $\mathbb{F}^n$.

Say I have functions $p_1,\ldots,p_m : \mathbb{F}^n \to \mathbb{F}$ each with high degree $d(p_i) \approx n$, and another function $s : \mathbb{F}^m \to \mathbb{F}$ also with high degree $d(s) \approx m$. Let $f : \mathbb{F}^n \to \mathbb{F}$ be the composition $f(x) := s(p_1(x),\ldots,p_n(x))$. Are there general properties of $s$ and the $p_i$ that characterize when $f$ can have very low degree, i.e. $d(f) \ll n$?

It's easy to come up with toy examples -- I'm wondering if there are any general statements one can make. E.g., some relationship among the $p_i$ that says that no choice of $s$ can reduce the degree too much, or a proof that for any set of $p_i$ there is some $s$ that reduces the degree by a lot, etc.

$\endgroup$
4
  • $\begingroup$ You talk about polynomials, but seem to be thinking of them as functions, which makes a difference over a finite field. Could you make it clear whether you really do mean polynomials or polynomial functions (in which case you might clarify what you mean by "degree")? $\endgroup$ Commented Aug 1, 2015 at 18:00
  • $\begingroup$ I don't think it was that unclear, but I modified it a bit to address your point. $\endgroup$
    – user94741
    Commented Aug 1, 2015 at 18:26
  • 3
    $\begingroup$ I think @JeremyRickard's point is that one usually doesn't speak of the degree of a polynomial function on a finite field, because such a function is induced by (infinitely) many different polynomials of (infinitely) many different degrees. Do you mean by the 'degree' of a polynomial function $f$ the least number $d$ such that $f$ is induced by a polynomial in which every monomial is a product of at most $d$ terms? $\endgroup$
    – LSpice
    Commented Aug 1, 2015 at 19:26
  • 1
    $\begingroup$ Thanks @JeremyRickard and L Spice for the feedback; I added some further clarification. $\endgroup$
    – user94741
    Commented Aug 3, 2015 at 20:24

0

You must log in to answer this question.

Browse other questions tagged .