9
$\begingroup$

Suppose that $f$ is a function from the prime-order field $\mathbb F_p$ to the field itself. Considering the evaluation map $P\mapsto P(x,f(x))$ and comparing the dimensions, it is easy to show that there exists a bivariate polynomial $P$ of total degree $d<\sqrt{2p}$ such that $P(x,f(x))=0$ for each $x\in\mathbb F_p$.

What I need instead is a low-degree polynomial $P$ such that $P(x,f(x))=0$ holds for all, but exactly one value of $x\in\mathbb F_p$.

In general, such a polynomial may fail to exist; say, to have $P(x,x)=0$ for all $x\in\mathbb F_p$ with one exception, we need $\deg P\ge p-1$.

What should one require from $f$ to ensure the existence of a polynomial $P$ with, say, $\deg P=O(p^{1/2+\varepsilon})$, such that $P$ vanishes on all, but exactly one point $(x,f(x))$, $x\in\mathbb F_p$?


To give the question a more precise shape, suppose that there is a partition $\mathbb F_p=S_1\cup\dotsb\cup S_n$, and a system of (univariate) polynomials $Q_1,\dotsc,Q_n$, such that $f$ coincides with $Q_k$ on $S_k$, for each $k\in[1,n]$. If in this case $P(x,f(x))$ vanishes for all, but one element $x\in\mathbb F_p$, then $$ \textstyle \deg P\ge \min \left\{ \frac{|S_k|-1)}{\deg Q_k}\colon 1\le k\le n \right\}. $$ Thus, a necessary condition for $P$ of degree $\deg P=O(p^{1/2+\varepsilon})$ to exist is that if $f$ is "piecewise polynomial", as we have just described, then $|S_k|=O(p^{1/2+\varepsilon}\deg Q_k)$ for each $k\in[1,n]$. Is some form of this condition also sufficient?

Making the even stronger assumption that $$ | \{ x\in\mathbb F_p\colon f(x)=Q(x) \} | = O(p^{1/2+\varepsilon} \deg Q) $$ for any (non-zero) polynomial $Q$, can we conclude that there exists a polynomial $P$ with $\deg P=O(p^{1/2+\varepsilon})$ such that $P$ vanishes on all, but exactly one point $(x,f(x)),\ x\in\mathbb F_p$?

$\endgroup$
3
  • $\begingroup$ This looks like list decoding. $\endgroup$ Aug 22, 2019 at 21:59
  • $\begingroup$ @FelipeVoloch: Could you expand on this? I cannot see any connection. $\endgroup$
    – Seva
    Aug 23, 2019 at 5:33
  • $\begingroup$ It may not be the same thing but it reminded me of Sudan's algorithm for list decoding of Reed-Solomon codes. $\endgroup$ Aug 23, 2019 at 15:35

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.