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

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

2.3. GREATEST COMMON DIVISORS 47<br />

93060801700<br />

1557792607653 x + 23315940650<br />

173088067517<br />

and, finally,<br />

761030000733847895048691<br />

.<br />

86603128130467228900<br />

Since this is a number, it follows that no polynomial can divide both A and B,<br />

i.e. that gcd(A, B) = 1.<br />

It is obvious that these calculations on polynomials with rational coefficients<br />

require several g.c.d. calculations on integers, and that the integers in these<br />

calculations are not always small.<br />

We can eliminate these g.c.d. calculations by working all the time with<br />

polynomials with integer coefficients, and this gives a generalisation <strong>of</strong> the a i <strong>of</strong><br />

algorithm 1, known as polynomial remainder sequences or p.r.s., by extending<br />

the definition <strong>of</strong> division.<br />

Definition 30 Instead <strong>of</strong> dividing f by g in K[x], we can multiply f by a<br />

suitable power <strong>of</strong> the leading coefficient <strong>of</strong> g, so that the divisions stay in R.<br />

The pseudo-remainder <strong>of</strong> dividing f by g, written prem(f, g), is the remainder<br />

when one divides lc(g) deg(f)−deg(g)+1 f by g, conceptually in K[x], but in fact all<br />

the calculations can be per<strong>for</strong>med in R, i.e. all divisions are exact in R. This<br />

is denoted 25 by prem(f, g).<br />

In some applications (section 3.1.9) it is necessary to keep track <strong>of</strong> the signs:<br />

we define a signed polynomial remainder sequence or s.p.r.s. <strong>of</strong> f 0 = f and<br />

f 1 = g to have f i proportional by a positive constant to −rem(f i−2 , f i−1 ).<br />

This gives us a pseudo-euclidean algorithm, analogous to algorithm 1 where we<br />

replace rem by prem, and fix up the contents afterwards. In the above example,<br />

we deduce the following sequence:<br />

and<br />

−15 x 4 + 381 x 2 − 261,<br />

6771195 x 2 − 30375 x − 4839345,<br />

500745295852028212500 x + 1129134141014747231250<br />

7436622422540486538114177255855890572956445312500.<br />

Again, this is a number, so gcd(A, B) = 1. We have eliminated the fractions,<br />

but at a cost <strong>of</strong> even larger numbers. Can we do better?<br />

2.3.2 Subresultant sequences<br />

One option would be to make the a i primitive at each step, since we are going<br />

to fix up the content terms later: giving the so-called primitive p.r.s. algorithm,<br />

which in this case would give<br />

−5x 4 + 127x 2 − 87; 5573x 2 − 25x − 3983; −4196868317x − 1861216034; 1<br />

25 This definition agrees with Maple, but not with all s<strong>of</strong>tware systems, which <strong>of</strong>ten use prem<br />

to denote what Maple calls sprem, i.e. only raising lc(g) to the smallest power necessary.

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

Saved successfully!

Ooh no, something went wrong!