19.07.2014 Views

Contents - Student subdomain for University of Bath

Contents - Student subdomain for University of Bath

Contents - Student subdomain for University of Bath

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

B.2. EQUALITY OF FACTORED POLYNOMIALS 223<br />

with the f i and g j square-free and relatively prime, i.e.:w gcd(f i , f i ′) = 1,<br />

gcd(g j , g j ′) = 1), is f ? =g.<br />

The obvious solution is to expand f and g, and check that the expanded <strong>for</strong>ms<br />

(which are canonical) are equal. Can we do better?<br />

One important preliminary remark is that the square-free representation <strong>of</strong><br />

a polynomial is unique. This leads to the following result.<br />

Proposition 78 If f = g, then every a i has to be a b j , and vice versa. For<br />

each such occurring value k, we must verify<br />

f k =<br />

n∏<br />

?<br />

f i =gk =<br />

i=1<br />

a i =k<br />

m∏<br />

g j .<br />

j=1<br />

b j =k<br />

(B.1)<br />

In particular, f k and g k must have the same degree, i.e.<br />

n∑<br />

deg(f i ) =<br />

i=1<br />

a i =k<br />

m∑<br />

deg(g j ).<br />

j=1<br />

b j =k<br />

(B.2)<br />

Again, we could check f k<br />

?<br />

=gk by expansion, but there is a better way.<br />

Example 16 Let f = x 2l −1 and g = (x−1)(x+1)(x 2 +1) · · · (x 2l−1 +1), where<br />

both are square-free, so proposition 78 does not help. f is already expanded, but<br />

expansion <strong>of</strong> g can give rise to x 2l −1 + x 2l −2 + · · · + x + 1, which has length<br />

O(2 l ), whereas g has length O(l).<br />

From now on we will assume that we are working over a domain that includes<br />

the integers.<br />

Lemma 14 If f and g have degree at most N, and agree at N + 1 different<br />

values, then f = g.<br />

Pro<strong>of</strong>. f − g is a polynomial <strong>of</strong> degree at most N, but has N + 1 zeros, which<br />

contradicts proposition 5 if it is non-zero.<br />

Hence it suffices to evaluate f and g at N + 1 points x i and check that the<br />

values agree. It is not necessary to construct any polynomial, and the integers<br />

involved are bounded (if we choose |x i | < N) by BN N , where B is a function<br />

<strong>of</strong> the coefficients <strong>of</strong> the f i , g j . Furthermore, we can evaluate at all these points<br />

in parallel. However, it does seem that we need to evaluate at N + 1 points, or<br />

very nearly so, even if f and g are very small.<br />

Open Problem 16 (Crossings <strong>of</strong> factored polynomials) Produce some nontrivial<br />

bounds on the maximum number <strong>of</strong> zeros <strong>of</strong> f − g, where f and g have<br />

small factored representations. See [RR90].<br />

The best we can say is as follows. Suppose, in the notation <strong>of</strong> (B.1), each f i<br />

has k i non-zero terms, and g j has l j non-zero terms, and no f i or g j is x (if

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!