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.

132 CHAPTER 4. MODULAR METHODS<br />

4. that we can use single-word primes;<br />

the running time <strong>of</strong> the ‘modular first’ (Figure 4.10) algorithm is<br />

O ( l 2 (d + 1) n + (d + 1) n+1) .<br />

This contrasts with the subresultant algorithm’s bound [op. cit. (80)] (assuming<br />

the p.r.s. are normal) <strong>of</strong><br />

(<br />

O l 2 (d + 1) 4n 2 2n2 3 n) .<br />

The real challenge comes with the potential sparsity <strong>of</strong> the polynomials.<br />

The factor (d + 1) n in the running time comes simply from the fact that a dense<br />

polynomial <strong>of</strong> degree d in n variables has that many terms, and we will need<br />

(d + 1) n−1 univariate g.c.d.s (there<strong>for</strong>e (d + 1) n coefficients) to deduce these<br />

potential terms.<br />

4.3.1 A worked example<br />

Let f be<br />

x 5 y 7 z 4 + x 4 y 8 z 4 + x 4 y 7 z 4 + x 3 y 7 z 4 + x 2 y 8 z 4 + x 5 y 3 z 5 + x 4 y 4 z 5 +<br />

x 3 y 4 z 6 + x 2 y 7 z 4 + x 2 y 5 z 6 − x 5 y 7 + x 5 yz 6 − x 4 y 8 + x 4 y 3 z 5 +<br />

x 4 y 2 z 6 − x 3 y 4 z 5 − x 2 y 5 z 5 + x 2 y 4 z 6 − x 4 y 7 + x 4 yz 6 − x 2 y 4 z 5 − x 3 y 7 −<br />

x 2 y 8 − x 2 y 7 − x 5 y 3 − x 4 y 4 − x 4 y 3 − x 5 y − x 4 y 2 − x 4 y + x 3 z+<br />

x 2 yz − x 3 − x 2 y + x 2 z − x 2 + xz + yz − x − y + z − 1<br />

(4.6)<br />

and g be<br />

x 6 y 7 z 4 − x 4 y 7 z 4 + x 6 y 3 z 5 + x 4 y 7 z 3 + x 4 y 4 z 6 − x 6 y 7 + x 6 yz 6 + x 4 y 7 z 2<br />

−x 4 y 4 z 5 − 2 x 2 y 7 z 4 + x 4 y 7 z − 2 x 4 y 3 z 5 + x 2 y 7 z 3 − 2 x 2 y 4 z 6 + 2 x 4 y 7 +<br />

x 4 y 3 z 4 − 2 x 4 yz 6 + x 2 y 7 z 2 + 3 x 2 y 4 z 5 + x 4 y 3 z 3 + x 4 yz 5 + x 2 y 7 z−<br />

x 6 y 3 + x 4 y 3 z 2 + x 4 yz 4 + 3 x 2 y 7 + x 4 y 3 z + x 4 yz 3 − x 6 y + 3 x 4 y 3 +<br />

x 4 yz 2 + x 4 yz + 3 x 4 y + x 4 z − x 4 − x 2 z + 2 x 2 − 2 z + 3<br />

(4.7)<br />

polynomials with 42 and 39 terms respectively (as against the 378 and 392 they<br />

would have if they were dense <strong>of</strong> the same degree).<br />

Let us regard x as the variable to be preserved throughout. If we compute<br />

gcd(f| z=2 , g| z=2 ) (which would require at least 4 8 y values) we get (assuming<br />

that z = 2 is a good reduction)<br />

gcd(f, g)| z=2 = ( 15 y 7 + 31 y 3 + 63 y ) x 4 + ( 15 y 7 + 32 y 4 + 1 ) x 2 + 1. (4.8)<br />

We note that each coefficient (with respect to x) has at most three terms.<br />

A dense algorithm would compute five more equivalents <strong>of</strong> (4.8), at different<br />

4 In fact, with y = 0 both f and g drop in x-degree, whereas with y = −2 we get a g.c.d.<br />

with x-degree 5, as y = −2 is a root <strong>of</strong> the resultant in Lemma 8.

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

Saved successfully!

Ooh no, something went wrong!