Access Full Complete Solution Manual Here

https://www.book4me.xyz/solution-manual-a-friendly-introduction-to-numerical-analysis-bradie/

Algorithms 1

Solutions

Chapter 1 Getting Started

1.1 Algorithms

1. Use the statistics algorithm from the text to compute the mean, x̄, and the
standard deviation, s, of the data set: −5, −3, 2, −2, 1.

The inputs are

x1 = −5, x2 = −3, x3 = 2, x4 = −2, x5 = 1, and n = 5.

Working sequentially through the steps of the algorithm, we find

STEP 1: xsum = 0; x2sum = 0

STEP 2: i = 1: xsum = 0 + (−5) = −5; x2sum = 0 + (−5)2 = 25
i = 2: xsum = −5 + (−3) = −8; x2sum = 25 + (−3)2 = 34
i = 3: xsum = −8 + 2 = −6; x2sum = 34 + 22 = 38
i = 4: xsum = −6 + (−2) = −8; x2sum = 38 + (−2)2 = 42
i = 5: xsum = −8 + 1 = −7; x2sum = 42 + 12 = 43
STEP 3: xbarp= −7/5 = −1.4
STEP 4: s = (5(43) − (−7)2 )/(5 · 4) = 2.880972
OUTPUT: xbar = −1.4 and s = 2.880972

Thus, for the data set consisting of the five numbers −5, −3, 2, −2, and 1, the
mean is x̄ = −1.4 and the standard deviation is s = 2.880972.

2. With n = 4, use the trapezoidal rule algorithm from the text to approximate
the value of the definite integral
1
1
Z
dx.
0 1 + x2

Rb
Matching this specific problem to the general pattern a f (x) dx, we see that a = 0,
1
b = 1 and f (x) = 1+x 2 . With n = 4, the trapezoidal rule algorithm yields
https://www.book4me.xyz/solution-manual-a-friendly-introduction-to-numerical-analysis-bradie/

2 Section 1.1

STEP 1: h = (1 − 0)/4 = 1/4; sum = 0

STEP 2: i = 1: sum = 0 + 1/(1 + (1/4)2 ) = 16/17
i = 2: sum = 16/17 + 1/(1 + (1/2)2 ) = 148/85
i = 3: sum = 148/85 + 1/(1 + (3/4)2 ) = 1012/425
STEP 3: sum = 2(1012/425) = 2024/425
STEP 4: sum = 2024/425 + 1/1 + 1/2 = 5323/850
OUTPUT: (1/8)(5323/850) = 5323/6800 = 0.782794117

Therefore,
1
1
Z
dx ≈ 0.782794117.
0 1 + x2
The exact value of the integral is π/4, so the absolute error in the trapezoidal rule
approximation is |π/4 − 0.782794117| ≈ 2.604 × 10−3 .
√
3. Use the square root algorithm from the text to approximate 5. Take x0 = 5,
ǫ = 5 × 10−4 and N max = 10.

Let a = 5, x0 = 5, ǫ = 5 × 10−4 , and N max = 10. The first iteration of the square
root algorithm yields

STEP 1: iter = 1
STEP 2: x1 = (5 + 5/5)/2 = 3
STEP 3: |x1 − x0 | = 2 > 5 × 10−4 , so
STEP 4: x0 = 3

In the second iteration, we find

STEP 1: iter = 2
STEP 2: x1 = (3 + 5/3)/2 = 7/3
STEP 3: |x1 − x0 | = 2/3 > 5 × 10−4 , so
STEP 4: x0 = 7/3

In the third iteration, we find

STEP 1: iter = 3
STEP 2: x1 = (7/3 + 15/7)/2 = 47/21
STEP 3: |x1 − x0 | = 2/21 > 5 × 10−4 , so
STEP 4: x0 = 47/21

The fourth iteration yields

STEP 1: iter = 4
STEP 2: x1 = (47/21 + 105/47)/2 = 2207/987
STEP 3: |x1 − x0 | = 2/987 > 5 × 10−4 , so
STEP 4: x0 = 2207/987

and the fifth iteration produces

https://www.book4me.xyz/solution-manual-a-friendly-introduction-to-numerical-analysis-bradie/

Algorithms 3

STEP 1: iter = 5
STEP 2: x1 = (2207/987 + 4935/2207)/2 = 4870847/2178309
STEP 3: |x1 − x0 | = 2/2178309 ≈ 9.18 × 10−7 < 5 × 10−4 , so
OUTPUT x1 = 4870847/2178309 ≈ 2.236067977
√
Thus, 5 ≈ 2.236067977.

4. A different scheme for approximating the square root of a positive real number
a is based on the recursive formula

x3n + 3xn a
xn+1 = .
3x2n + a

(a) Construct an algorithm for approximating the square root of a given posi-
tive real number a using this formula.
(b) Test your algorithm using a = 2 and x0 = 2. Allow a maximum of 10
iterations and use a convergence tolerance of ǫ = 5 × 10−5 . Compare the
performance of this algorithm with the one presented in the text.

(a) GIVEN: nonnegative real number a

starting approximation x0
convergence parameter ǫ
maximum number of iterations N max

STEP 1: for iter from 1 to Nmax

STEP 2: compute x1 = (x30 + 3ax0 )/(3x20 + a)
STEP 3: if |x1 − x0 | < ǫ, OUTPUT x1
STEP 4: copy the value of x1 to x0
end
OUTPUT: “maximum number of iterations exceeded”
(b) Let a = 2, x0 = 2, ǫ = 5 × 10−5 , and N max = 10. The first iteration of the
algorithm from part (a) yields
STEP 1: iter = 1
STEP 2: x1 = (8 + 12)/(12 + 2) = 10/7
STEP 3: |x1 − x0 | = 4/7 > 5 × 10−5 , so
STEP 4: x0 = 10/7
In the second iteration, we find
STEP 1: iter = 2
STEP 2: x1 = (1000/343 + 60/7)/(300/49 + 2) = 3940/2786
STEP 3: |x1 − x0 | = 40/2786 > 5 × 10−5 , so
STEP 4: x0 = 3940/2786
The third iteration then yields
https://www.book4me.xyz/solution-manual-a-friendly-introduction-to-numerical-analysis-bradie/

4 Section 1.1

STEP 1: iter = 3
STEP 2: x1 = 1.414213562
STEP 3: |x1 − x0 | ≈ 3.64 × 10−7 < 5 × 10−5 , so
OUTPUT x1 = 1.414213562
√
Thus, 2 ≈ 1.414213562, which is correct to the digits shown. Observe that
with the same number of iterations (three), the current algorithm produced a
more accurate approximation than the algorithm from the text.

5. Let A be an n × m matrix and B be an m × p matrix. The n × p matrix C = AB

has elements defined by
Xm
cik = aij bjk
j=1

for each i = 1, 2, 3, . . . , n and each k = 1, 2, 3, . . . , p. Construct an algorithm to

compute the product of two matrices.

GIVEN: the integers n, m and p

the elements in the matrix A, aij
the elements in the matrix B, bjk

STEP 1: for i from 1 to n

STEP 2: for k from 1 to p
STEP 3: initialize sum to 0
STEP 4: for j from 1 to m
add aij bjk to sum
end
STEP 5: set cik = sum
end
end
OUTPUT: the elements in the matrix C, cik

6. Consider the computation of the following sum,

n X
X n
ai bj ,
i=1 j=1

where the ai and bj are real numbers.

(a) How many multiplications and how many additions are required to com-
pute the sum? Each answer should be a function of n.
(b) Modify the summation to an equivalent form which reduces the number
of operations needed. How many multiplications and how many additions
are required to compute the sum in its revised form?
https://www.book4me.xyz/solution-manual-a-friendly-introduction-to-numerical-analysis-bradie/

Algorithms 5

(a) As written, forming the product of ai and bj for each possible combination
of i and j requires n2 multiplications. Summing all n2 terms requires n2 − 1
additions.
(b) Observe that
n
n X n
! n 
X X X
ai bj = ai  bj  .
i=1 j=1 i=1 j=1

Using the expression on the right-hand side, the original sum can now be
computed using only 2n − 2 additions and a single multiplication.

7. Let a be a non-zero real number. For any x0 satisfying 0 < x0 < 2/a, the
recursive sequence defined by
xn+1 = xn (2 − axn )
converges to 1/a.
(a) Construct an algorithm for approximating the reciprocal of a given non-
zero real number a using this formula.
(b) Test your algorithm using a = 37 and x0 = 0.01. Allow a maximum of 10
iterations and use a convergence tolerance of ǫ = 5 × 10−4 .

(a) GIVEN: non-zero real number a

starting approximation x0
convergence parameter ǫ
maximum number of iterations Nmax

STEP 1: for iter from 1 to Nmax

STEP 2: compute x1 = x0 (2 − ax0 )
STEP 3: if |x1 − x0 | < ǫ, OUTPUT x1
STEP 4: copy the value of x1 to x0
end
OUTPUT: “maximum number of iterations has been exceeded”
(b) Let a = 37, x0 = 0.01, ǫ = 5 × 10−4 , and N max = 10. The first iteration of
the algorithm from part (a) yields
STEP 1: iter = 1
STEP 2: x1 = 0.01(2 − 0.01 · 37) = 0.0163
STEP 3: |x1 − x0 | = 0.0063 > 5 × 10−4 , so
STEP 4: x0 = 0.0163
In the second iteration, we find
STEP 1: iter = 2
STEP 2: x1 = 0.0163(2 − 0.0163 · 37) = 0.02276947
STEP 3: |x1 − x0 | = 0.00646947 > 5 × 10−4 , so
STEP 4: x0 = 0.02276947
https://www.book4me.xyz/solution-manual-a-friendly-introduction-to-numerical-analysis-bradie/

6 Section 1.1

The third iteration then yields

STEP 1: iter = 3
STEP 2: x1 = 0.026356335
STEP 3: |x1 − x0 | ≈ 3.59 × 10−3 > 5 × 10−4 , so
STEP 4: x0 = 0.026356335
In the fourth iteration, we calculate
STEP 1: iter = 4
STEP 2: x1 = 0.027010383
STEP 3: |x1 − x0 | ≈ 6.54 × 10−4 > 5 × 10−4 , so
STEP 4: x0 = 0.027010383
With the fifth iteration, we find
STEP 1: iter = 5
STEP 2: x1 = 0.027027016
STEP 3: |x1 − x0 | ≈ 1.66 × 10−5 < 5 × 10−4 , so
OUTPUT x1 = 0.027027016
Thus, 1/37 ≈ 0.027027016, which is in error by roughly 1.10 × 10−7 .

8. Given two positive integers a and b, the greatest common divisor of a and b
is the largest integer which divides both a and b; i.e., the largest integer n for
which both a/n and b/n are integers.
(a) Construct an algorithm to compute the greatest common divisor of two
positive integers.
(b) How many divisions does your algorithm require?

(a) GIVEN: positive integers a and b

STEP 1: initialize c = min(a, b) and gcd = 1

STEP 2: for n from 2 to c
STEP 3: if a/n and b/n are integers, set gcd = n
OUTPUT: gcd
(b) The above algorithm requires 2(c − 1) divisions, where c = min(a, b).

9. The inner product, or dot product, of two n-vectors x and y is given by

x · y = x1 y1 + x2 y2 + x3 y3 + · · · + xn yn .

(a) Construct an algorithm to compute the inner product of two n-vectors.

(b) Apply your algorithm to calculate the inner product of the vectors
 T  T
x= −3 4 1 2 and y= 1 −3 2 5
https://www.book4me.xyz/solution-manual-a-friendly-introduction-to-numerical-analysis-bradie/

Algorithms 7

(a) GIVEN: n-vectors x and y

vector dimension n

STEP 1: initialize sum = 0

STEP 2: for i from 1 to n
add xi · yi to sum
OUTPUT: sum
(b) The inputs are the vectors

x = [ −3 4 1 2 ] and y = [ 1 −3 2 5 ]

and n = 4. Working sequentially through the steps of the algorithm, we find

STEP 1: sum = 0
STEP 2: i = 1: sum = 0 + (−5)(1) = −3
i = 2: sum = −3 + (4)(−3) = −15
i = 3: sum = −15 + (1)(2) = −13
i = 4: sum = −13 + (2)(5) = −3
OUTPUT: sum = −3
Thus, x · y = −3.

10. The linear correlation coefficient for n ordered pairs (xi , yi ) is given by the
formula
Pn Pn Pn
n i=1 xi yi − ( i=1 xi ) ( i=1 yi )
r=q P q .
n Pn 2 Pn Pn 2
n i=1 x2i − ( i=1 xi ) n i=1 yi2 − ( i=1 yi )

(a) Construct an algorithm to compute the linear correlation coefficient for a

given set of ordered pairs.
(b) Apply your algorithm to compute the linear correlation coefficient for the
following set of ordered pairs:
xi 3 7 9 2 7 0 3
yi -5 10 15 -8 11 -10 -4

(a) GIVEN: ordered pairs (xi , yi )

number of ordered pairs n

STEP 1: initialize xsum, ysum, x2sum, y2sum and xysum to 0

STEP 2: for i from 1 to n
add xi to xsum
add yi to ysum
add x2i to x2sum
add yi2 to y2sum
add xi · yi to xysum
https://www.book4me.xyz/solution-manual-a-friendly-introduction-to-numerical-analysis-bradie/

8 Section 1.1

end
n · xysum − (xsum)(ysum)
STEP 3: calculate r = p p
n · x2sum − (xsum)2 · n · y2sum − (ysum)2 )
OUTPUT: r
(b) Working sequentially through the steps of the algorithm, we find

STEP 1: xsum = ysum = x2sum = y2sum = xysum = 0

STEP 2: i = 1: xsum = 0 + 3 = 3; ysum = 0 + (−5) = −5
x2sum = 0 + 32 = 9; y2sum = 0 + (−5)2 = 25
xysum = 0 + (3)(−5) = −15
i = 2: xsum = 3 + 7 = 10; ysum = −5 + 10 = 5
x2sum = 9 + 72 = 58; y2sum = 25 + 102 = 125
xysum = −15 + (7)(10) = 55
i = 3: xsum = 10 + 9 = 19; ysum = 5 + 15 = 20
x2sum = 58 + 92 = 139; y2sum = 125 + 152 = 350
xysum = 55 + 9(15) = 190
i = 4: xsum = 19 + 2 = 21; ysum = 20 + (−8) = 12
x2sum = 139 + 22 = 143; y2sum = 350 + (−8)2 = 414
xysum = 190 + (2)(−8) = 174
i = 5: xsum = 21 + 7 = 28; ysum = 12 + 11 = 23
x2sum = 143 + 72 = 192; y2sum = 414 + 112 = 535
xysum = 174 + (7)(11) = 251
i = 6: xsum = 28 + 0 = 28; ysum = 23 + (−10) = 13
x2sum = 192 + 0= 192; y2sum = 535 + (−10)2 = 635
xysum = 251 + (0)(−10) = 251
i = 7: xsum = 28 + 3 = 31; ysum = 13 + (−4) = 9
x2sum = 192 + 32 = 201; y2sum = 635 + (−4)2 = 651
xysum = 251 + (3)(−4) = 239

7(239) − (31)(9)
STEP 3: r= p p = 0.987
7(201) − (31)2 · 7(651) − (9)2 )

OUTPUT: r = 0.987

Thus, the linear correlation coefficient of the seven given data pairs is 0.987.

11. The midpoint rule approximates the value of a definite integral using the formula
Z b n
X
f (x)dx ≈ 2h f (xj ),
a j=1

where h = (b − a)/2n and xj = a + (2j − 1)h.

(a) Construct an algorithm to approximate the value of a definite integral using
the midpoint rule.
https://www.book4me.xyz/solution-manual-a-friendly-introduction-to-numerical-analysis-bradie/

Algorithms 9

R2
(b) Apply your algorithm to approximate the value of 1 dx/x. Take n = 4.
Compare the approximation obtained from the midpoint rule with the
approximation obtained from the trapezoidal rule.

(a) GIVEN: the limits of integration a and b

the integrand f
the number of subintervals n

STEP 1: compute h = (b − a)/(2n); initialize sum = 0

STEP 2: for j from 1 to n
add f (a + (2j − 1)h) to sum
OUTPUT: 2h · sum
Rb
(b) Matching this specific problem to the general pattern a f (x) dx, we see that
a = 1, b = 2 and f (x) = x1 . With n = 4, the midpoint rule algorithm yields
STEP 1: h = (2 − 1)/(2 · 4) = 1/8; sum = 0
STEP 2: j = 1: sum = 0 + 1/(1 + 1/8) = 8/9
j = 2: sum = 8/9 + 1/(1 + 3/8) = 160/99
j = 3: sum = 160/99 + 1/(1 + 5/8) = 2872/1287
j = 4: sum = 2872/1287 + 1/(1 + 7/8) = 17792/6435
OUTPUT: 2(1/8)(17792/6435) = 4448/6435 = 0.691219891
Therefore,
2
1
Z
dx ≈ 0.691219891.
1 x
The exact value of the integral is ln 2, so the absolute error in the midpoint rule
approximation is | ln 2 − 0.691219891| ≈ 1.927 × 10−3 . This error is roughly
half the error in the trapezoidal rule approximation obtained in the text.

12. Consider the following algorithm for the trapezoidal rule:

GIVEN: the limits of integration a and b

the integrand f
the number of subintervals n
STEP 1: compute h = (b − a)/n; and initialize sum to 0
STEP 2: for i from 1 to n − 1
add 2f (a + ih) to sum
STEP 3: add f (a) and f (b) to sum
OUTPUT: (h/2)sum

Compare the number of arithmetic operations required by this algorithm to the

number of operations required by the algorithm presented in the text.

Aside from the operations needed to evaluate f , the trapezoidal rule algorithm
presented in the text uses 2n + 1 additions/subtractions (one in STEP 1, 2(n − 1)
https://www.book4me.xyz/solution-manual-a-friendly-introduction-to-numerical-analysis-bradie/

10 Section 1.1

in STEP 2 and two in STEP 4) and n + 3 multiplications/divisions (one in STEP 1,

n − 1 in STEP 2, one in STEP 3 and two in the OUTPUT). The modified algorithm
presented in this exercise uses the exact same additions/subtractions, but requires
a total of 2n + 1 multiplications/divisions (one in STEP 1, 2(n − 1) in STEP 2 and
2 in the OUTPUT).

13. Rewrite the algorithm for the trapezoidal rule which was presented in the text to
reduce both the number of additions and the number of multiplications/divisions
by one.

The key is to observe that

" n−1
# " n−1
#
h X f (a) + f (b) X
f (a) + 2 f (xi ) + f (b) = h + f (xi ) .
2 i=1
2 i=1

This suggests modifying the trapezoidal rule algorithm to

GIVEN: the limits of integration a and b
the integrand f
the number of subintervals n

STEP 1: compute h = (b − a)/n; initialize sum = (f (a) + f (b))/2

STEP 2: for i from 1 to n-1
add f (a + ih) to sum
OUTPUT: h · sum
This algorithm uses 2n additions/subtractions (two in STEP 1 and 2(n − 1) in
STEP 2) and n + 2 multiplications/divisions (two in STEP 1, n − 1 in STEP 2 and
one in the OUTPUT).

14. Let P (x) = an xn + an−1 xn−1 + · · · + a1 x + a0 be an n-th degree polynomial

with all real coefficients, and let x0 be a given real number.
(a) Treating integer powers as repeated multiplication, how many multiplica-
tions and how many additions are required to evaluate P (x) at x = x0 ?
(b) Devise an algorithm for computing the value of an n-th degree polynomial
which reduces the required number of arithmetic operations? How many
multiplications and how many additions are required by your algorithm?

(a) For each j, evaluation of the term aj xj0 requires j multiplications (j − 1 to

evaluate the power of x0 and one more to include the coefficient). Thus, to
evaluate the entire polynomial requires
n
X n(n + 1)
j=
j=1
2
https://www.book4me.xyz/solution-manual-a-friendly-introduction-to-numerical-analysis-bradie/

Algorithms 11

multiplications. Because the polynomial consists of n + 1 terms, a total of n

additions are needed for evaluation.
(b) To devise a more efficient scheme for evaluating an nth-degree polynomial, we
rewrite the polynomial in nested form:

P (x) = a0 + x(a1 + a2 x + a3 x2 + · · · an xn−1 )

= a0 + x(a1 + x(a2 + a3 x + · · · + an xn−2 ))
= ···
= a0 + x(a1 + x(a2 + x(a3 + · · · x(an−1 + an x) · · ·))).

Based on this form, we can evaluate the polynomial with the following algo-
rithm:
GIVEN: the polynomial coefficients a0 , a1 , a2 , ..., an
value of independent variable x0
the degree of the polynomial n

STEP 1: initialize value = an

STEP 2: for i from n − 1 downto 0
replace value by value · x0 + ai
OUTPUT: value
With this algorithm, polynomial evaluation can be carried out with n additions
and n multiplications.

For Exercises 15 - 18, make use of the fact that when the sum of a convergent
alternating series is approximated using the sum of the first n terms, the error
in P
this approximation is smaller than the magnitude of the (n+1)-st term; i.e.,
if (−1)n an is an alternating series with sum S, then
 
 n−1
X 
k
S − (−1) ak  < an .
 
 
k=0

15. The value of π is given by

∞
(−1)n
 
X 1 1 1
π=4 = 4 1 − + − + −··· .
n=0
2n + 1 3 5 7

(a) Construct an algorithm to approximate the value of π to within a specified

tolerance, ǫ.
(b) Test your algorithm with a tolerance value of ǫ = 5 × 10−3 .
https://www.book4me.xyz/solution-manual-a-friendly-introduction-to-numerical-analysis-bradie/

12 Section 1.1

(a) GIVEN: convergence parameter ǫ

maximum number of terms Nmax

STEP 1: initialize sum to 4 and sign to -1

STEP 2: for i from 1 to Nmax
STEP 3: set term = 4/(2i + 1)
STEP 4: if term < ǫ, OUTPUT sum
STEP 5: replace sum by sum + sign ∗ term
STEP 6: replace sign by −sign
end
OUTPUT: “more terms needed”
(b) With a tolerance of 5 × 10−3 , it takes 400 terms to produce the estimate
π ≈ 3.1391, which is in error by roughly 2.5 × 10−3 .

16. The value of 1/e is given by

∞
X (−1)n 1 1 1
1/e = = 1 − + − + −···.
n=0
n! 1! 2! 3!

(a) Construct an algorithm to approximate the value of 1/e to within a speci-

fied tolerance, ǫ.
(b) Test your algorithm with a tolerance value of ǫ = 5 × 10−7 .

(a) GIVEN: convergence parameter ǫ

maximum number of terms Nmax

STEP 1: initialize sum to 1, term to 1 and sign to -1

STEP 2: for i from 1 to Nmax
STEP 3: replace term by term/i
STEP 4: if term < ǫ, OUTPUT sum
STEP 5: replace sum by sum + sign ∗ term
STEP 6: replace sign by −sign
end
OUTPUT: “more terms needed”
(b) With a tolerance of 5 × 10−7 , ten terms are needed to produce the estimate
1/e ≈ 0.367879189, which is in error by roughly 2.5 × 10−7 .

17. The value of sin(π/10) is given by

∞
π X (−1)n  π 2n+1 π 1  π 3 1  π 5 1  π  7
sin = = − + − +− · · · .
10 n=0
(2n + 1)! 10 10 3! 10 5! 10 7! 10

(a) Construct an algorithm to approximate the value of sin(π/10) to within a

specified tolerance, ǫ.
https://www.book4me.xyz/solution-manual-a-friendly-introduction-to-numerical-analysis-bradie/

Algorithms 13

(b) Test your algorithm with a tolerance

√ value of ǫ = 5 × 10−7 . Note that the
1
exact value of sin(π/10) is 4 ( 5 − 1).

(a) GIVEN: convergence parameter ǫ

maximum number of terms Nmax

STEP 1: initialize sum to π/10, term to π/10 and sign to -1

STEP 2: for i from 1 to Nmax
STEP 3: replace term by term(π/10)2 /((2i)(2i + 1))
STEP 4: if term < ǫ, OUTPUT sum
STEP 5: replace sum by sum + sign ∗ term
STEP 6: replace sign by −sign
end
OUTPUT: “more terms needed”
(b) With a tolerance of 5 × 10−7 , three terms are needed to produce the estimate
sin(π/10) ≈ 0.309017054, which is in error by roughly 6.0 × 10−8 .

18. The value of cos(π/5) is given by

∞
π X (−1)n  π 2n 1  π 2 1  π 4 1  π 6
cos = =1− + − + −···.
5 n=0
(2n)! 5 2! 5 4! 5 6! 5

(a) Construct an algorithm to approximate the value of cos(π/5) to within a

specified tolerance, ǫ.
(b) Test your algorithm with a tolerance
√ value of ǫ = 5 × 10−7 . Note that the
1
exact value of cos(π/5) is 4 ( 5 + 1).

(a) GIVEN: convergence parameter ǫ

maximum number of terms Nmax

STEP 1: initialize sum to 1, term to 1 and sign to -1

STEP 2: for i from 1 to Nmax
STEP 3: replace term by term(π/5)2 /((2i − 1)(2i))
STEP 4: if term < ǫ, OUTPUT sum
STEP 5: replace sum by sum + sign ∗ term
STEP 6: replace sign by −sign
end
OUTPUT: “more terms needed”
(b) With a tolerance of 5 × 10−7 , five terms are needed to produce the estimate
cos(π/5) ≈ 0.809016997, which is in error by roughly 2.6 × 10−9 .