You are on page 1of 1

Numerical Analysis 2, Exercises for lecture 2 [§3.

1–4]
1* Find a Taylor polynomial approximation of 2 sinh x that ensures 6 correct
1 1
decimals for − 10 ≤ x ≤ 10 .

2* The power series



X xk
k=1 k
converges for −1 ≤ x < 1. Show that a partial sum with 9 terms
1 1
ensures 10 correct decimals for − 10 ≤ x ≤ 10 .

3* The zeroth order Bessel function of the first kind is given by the power
series

X (−1)k x2k
J0 (x) = 2k 2
k=0 2 (k!)

How many terms of the partial sum are needed to ensure 9 correct
1 1
decimals for − 10 ≤ x ≤ 10 ?

4** Encke’s method for computing the orbit of a satellite requires the eval-
uation of the function
1 − (1 − 2x)−3/2
f (x) =
x
Find a series representation that allows the function to be computed
accurately for small x. How many terms of the partial sum will give
1 1
an approximation with 4 correct decimals for − 100 ≤ x ≤ 100 ?

5** If range reduction is done in IEEE double precision (µ = 2−53 ), in how


large an interval around x = nπ might the relative error of sin(x) due
to range reduction be larger than 2−24 ? [page 63 exercise E2]

Answers
1 |x|5
1. p(x) = 2x + x2 /3, | sinh x − p(x)| ≤ < 0.5 · 10−6
60 1 − x2 /42
9
xk Z ∞ xk x9
= 0.48 · 10−10
X
2. f (x) − ≤ dk = −
k=1 k 9 9 9 ln(x)

x2 x4 x6
!
3. 3 terms, i.e. |J0 (x) − 1 − + |≤ = 0.4 · 10−9 .
4 64 2304
3·5 3·5·7 2 15 35
 
4. f (x) = −3− x− x −· · ·; three terms, i.e. |f (x)− −3 − x − x2 | ≤
2! 3! 2 2
945 |x|3 −4
≤ 0.4 · 10 .
24 1 − 11|x|/5
5. x = nπ(1 ± 2−28 )

You might also like