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.

Chapter 1<br />

Introduction<br />

Computer algebra, the use <strong>of</strong> computers to do algebra rather than simply arithmetic,<br />

is almost as old as computing itself, with the first theses [Kah53, Nol53]<br />

dating back to 1953. Indeed it was anticipated from the time <strong>of</strong> Babbage, when<br />

Ada Augusta, Countess <strong>of</strong> Lovelace, wrote<br />

We may say most aptly that the Analytical Engine weaves algebraical<br />

patterns just as the Jacquard loom weaves flowers and leaves.<br />

[Ada43]<br />

In fact, it is not even algebra <strong>for</strong> which we need s<strong>of</strong>tware packages, computers<br />

by themselves can’t actually do arithmetic: only a limited subset <strong>of</strong> it. If we ask<br />

Excel 1 to compute e π√ 163 −262537412640768744, we will be told that the answer<br />

is 256. More mysteriously, if we go back and look at the <strong>for</strong>mula in the cell, we<br />

see that it is now e π√ 163 −262537412640768800. In fact, 262537412640768744 is<br />

too large a whole number (or integer, as mathematicians say) <strong>for</strong> Excel to handle,<br />

and it has converted it into floating-point (what Excel terms “scientific”)<br />

notation. Excel, or any other s<strong>of</strong>tware using the IEEE standard [IEE85] representation<br />

<strong>for</strong> floating-point numbers, can only store them to a given accuracy,<br />

about 2 16 decimal places. 3 In fact, it requires twice this precision to show that<br />

e π√ 163 ≠ 262537412640768744. Since e π√ 163 = (−1) √ −163 , it follows from deep<br />

results <strong>of</strong> transcendental number theory [Bak75], that not only is e π√ 163 not<br />

an integer, it is not a fraction (or rational number), nor even the solution <strong>of</strong> a<br />

polynomial equation with integer coefficients: essentially it is a ‘new’ number.<br />

1 Or any similar s<strong>of</strong>tware package.<br />

2 We say ‘about’ since the internal representation is binary, rather than decimal.<br />

3 In fact, Excel is more complicated even than this, as the calculations in this table show.<br />

i 1 2 3 4 . . . 10 11 12 . . . 15 16<br />

a 10 i 10 100 1000 1. . . 0 . . . 1. . . 0 10 11 10 12 . . . 10 15 10 16<br />

b a-1 9 99 999 9999 9. . . 9 . . . 9. . . 9 10 12 . . . 10 15 10 16<br />

c a-b 1 1 1 1 1 . . . 1 1 . . . 1 0<br />

We can see that the printing changes at 12 decimal digits, but that actual accuracy is not lost<br />

until we subtract 1 from 10 16 .<br />

13

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

Saved successfully!

Ooh no, something went wrong!