A method to obtain the best uniform polynomial approximation...
Transcript of A method to obtain the best uniform polynomial approximation...
A method to obtain the best uniform polynomial
approximation for the family of rational
functioncbxax 2
1
M. A. Fariborzi Araghi1 ,F. Froozanfar2
1Department of Mathematics, Islamic Azad university, Central Tehran branch,
P.O.Box 13.185.768, Tehran, Iran.
2Ms.student of Mathematics, Islamic Azad university, Kermanshah branch
, Kermanshah, Iran
*Correspondence E‐mail: M. A. Fariborzi Araghi: [email protected]
© 2015 Copyright by Islamic Azad University, Rasht Branch, Rasht, Iran
Online version is available on: www.ijo.iaurasht.ac.ir
Abstract
In this article, by using Chebyshev’s polynomials and Chebyshev’s expansion,
we obtain the best uniform polynomial approximation out of nP2 to a class of
rational functions of theform 12 cax on any non symmetric interval ed , .
Using the obtained approximation, we provide the best uniform polynomial
approximation to a class of rational functions of the form 12 cbxax for
both cases 042 acb and 042 acb .
Key words: Chebyshev’s polynomials, Chebyshev’s expansion, uniform norm,
the best uniform polynomial approximation, alternating set.
1. Introduction
.
cos21
sinsincoscoscoscos12
1
0
ptt
ppntppntpntptxTtb
pp
npnnpnpnpn
j
pj
pj
In
section 2, we characterize the best On of the important and applicable subjects in
applied mathematics is the best approximation for functions. A large number of
paper and books have considered this problem in various points of view.
Volume 7, issue 1, Winter 2015, 753-766 Research Paper
Iranian Journal of Optimization, Vol 7, Issue 1,winter 2015 754
Definition 1. [19] Suppose nP denotes the space of polynomials of degree at most
n , then for given edCf , , there exists a unique polynomial nn Pp * such that:
.,*
nn Pppfpf
We call *
np the best polynomial approximation out of nP to f on ed , .
In other words, nn Pp * is the best uniform approximation for function f on
ed , if .,maxmax,, *
nexd
nexd
n PpxpxfxpxfedfE
The main questions of this problem are existence, uniqueness and characterization
of the solution. The existence and uniqueness of the solution for the best
approximation problem have been proved in [15,19].
In resent years, some researches investigated in order to characterize the best
uniform approximation for special classes of functions. Several of these
researches were focused on classes of functions possessing a certain expansion by
Chebyshev’s polynomials. For example Jokar and Mehri in [8] studied
11
aax and 1
1
x . Also Achieser in [1,2] studied 1 ax . Lorentz in
[10] obtained the best uniform approximation for complex
function 1z , Cz , . In the sequel, Lubinesky in [11] showed that
Lagrange interpolants at the Chebyshev zeros yield the best relevant polynomial
approximation of 121
ax on 1,1 . Eslahchi and Dehghan in [6] characterized
the best uniform polynomials approximation to a class of functions 122 xa on
1,1 and cc, . They also in [5] obtained the best uniform approximation to a
class of 1 xTaT qq . Also Elliott in [9] used the generalized form of
Chebyshev’s polynomials in a specific series to obtain the best approximation.
At first some definitions and theorems that will be used throughout this article are
introduced.
Theorem 1. (Chebyshev’s alternation theorem)[15]
Let f be in ],[ edC . Let the polynomial p be in nP and xpxfxe n . Then
p is the best uniform approximation
np to f on ed , if and only if there exist at
least 2n points 221 nxxx in ed , , for which:[14]
iniexd
i xpxfxe
max , with ii xexe 1 .
Iranian Journal of Optimization, Vol 7, Issue 1,winter 2015 755
Definition 2. [4,16] The Chebyshev’s polynomial in 1,1 of degree n is denoted
by nT and is defined by nxTn cos where xcos .
1
Note that nT is a polynomial of degree n with leading coefficient 12 n .
Definition 3. [12] The Chebyshev’s polynomial in ed , of degree n is denoted by *
nT and is defined by nxTn cos* where
de
edx
2cos .
2
Lemma 1. [8] For cosx , 1t and natural number p we have:
,
cos21
cos12
0
ptt
ptxTta
pp
p
j
pj
pj
uniform approximation to the class of 12 cax on ed , and in section 3, using
the results from section 2, we obtain the best uniform approximation for the class
of 12 cbxax on 1,1 .
2. Best Approximation of 12 cax on ed ,
In this section, we determine the best uniform polynomial approximation out of
nP2 to 12 cax on ed , , when 0
a
c or 0
a
c.
Now, we prove the following lemmas to verify Chebyshev’s expansion in two
mentioned cases.
Lemma 2. Suppose that edx , , 0a
c and 2
4de
a
c
. Then, we have:
3
);(
1
32
1
16112
0
2
42
2
42
2
22
xTttdea
t
tdea
t
xa
ca
caxk
k
k
Iranian Journal of Optimization, Vol 7, Issue 1,winter 2015 756
where 1,)(4
2)(
1 2
tde
a
c
a
c
det
4
and nxTn cos)( wherede
x
2cos .
Proof: In the expansion of the function 1,1 2
22
x on 1,1 we have [17]:
),(1
8
1
412
0
2
4
2
4
2
22xTt
t
t
t
t
xk
k
k
5
where, cosx andt
t
2
12 and 12 t . Suppose that
a
c , then
we have
).(1
8
1
4
cos2
1
12
0
2
4
2
4
2
2
22
xTtt
t
t
t
t
tk
k
k
6
According to 2 we can write:
.
22
14
4
cos2
1
1122
22
2
2
222
ed
xt
tde
de
t
txa
c
7
Combining 6 and 7 we obtain:
).()1(
32
)1(
16
22
1
42
0
2
42
2
42
2
222
2
xTttde
t
tde
t
edx
t
tde
k
k
k
8
Iranian Journal of Optimization, Vol 7, Issue 1,winter 2015 757
Suppose that
2)(
42
)(
1de
a
c
a
c
det , consequently 1t .( Note
that for )(
)(42 22
de
deaat
, the condition 1t is not true.)
Thus we have:
).(
)1(
32
)1(
16
2
42
0
2
42
2
42
2
2xTt
tde
t
tde
t
edx
a
ck
k
k
9
where with nxTn cos)( , de
x
2cos , so relation 3 is proved.
Lemma 3. Suppose that edx , and 0a
c. Then we have:
);(1
1
32
1
16112
0
2
42
2
42
2
22
xTttdea
t
tdea
t
a
cxa
caxk
k
kk
10
where
.1,2
41 2
t
a
cde
a
c
det
11
and nxTn cos)( wherede
x
2cos .
Proof: In the expansion of the function22
1
x on 1,1 we have[6]:
,1
1
8
1
41 *
2
0
2
4
2
4
2
22xTt
t
t
t
t
xk
k
kk
12
Iranian Journal of Optimization, Vol 7, Issue 1,winter 2015 758
where cosx ,t
t
2
1 2 .With suppose
ac ,the rest of proof is similar to
the proof of lemma 2. Thus we omit it.
Theorem 2. The best uniform polynomial approximation out of nP2
for 12 cax where 0
a
c, on ed , and 2
4de
a
c
, is as follows:
),(1
32)(
1
32
1
162242
22
2
1
0
2
42
2
42
2*
2 xTtdea
txTt
tdea
t
tdea
txp n
n
k
n
k
k
n
13
and 242
42*
22
1
32
tdea
tpffE
n
nn ,
14 where 1,)(4
2)(
1 2
tde
a
c
a
c
det , nxTn cos)(
wherede
x
2cos .
Proof: Noting to Chebyshev’s alternation theorem we should prove that the
function )(1
)( 222 xpcax
xe nn
15
has 22 n alternating points in ed , . From 3 and 13 we have:
).(1
32)(
1
322242
22
2
2
42
2
2 xTtdea
txTt
tdea
txe n
n
k
nk
k
n
16
By replacing 2p in lemma 1 and subtracting both sides of b from a we
obtain:
.2cos21
2sin2sin2cos2cos2cos)(
24
22
2
2
tt
nntntxTt n
k
nk
k
17
By replacing 17 in 16 , we obtain:
Iranian Journal of Optimization, Vol 7, Issue 1,winter 2015 759
ntt
tt
ntt
tttt
tdea
txe
n
n
2sin2cos21
2sin1
2cos2cos21
2cos2112cos1
1
32
24
42
24
2442
242
22
2
18
Noting to 4 , we have 2
2
4
1
t
tde
a
c. Then we can rewrite 18 in the
form of:
.2sin
44
2
2cos
8
1
32
22
4222
2
2
22
222
242
42
2
n
xa
cde
xdexdea
c
a
c
n
xa
cde
dea
cx
a
cde
tdea
txe
n
n
19
Now if we define:
20,
8
22
222
1
xa
cde
dea
cx
a
cde
xF
21.
44
2
22
4222
2
2
2
xa
cde
xdexdea
c
a
c
xF
Then we have:
.
282
2
22
2
2
2
1
xa
cde
dea
c
a
cx
xF
22
Iranian Journal of Optimization, Vol 7, Issue 1,winter 2015 760
It is easy to conclude that if ed,0 , then xF1 is a monotonic function and if
ed,0 then xF1 is a monotonic function on each interval 0,d , e,0 . Also we
have: .,,1)()( 2
2
2
1 edxxFxF
23
Therefore, according to 22 and definition of nT for edx , , we have:
1)(1 1 xF . Hence according to mean value theorem for every edx , , there
exists a ),0( such that .,),(cos 1 edxxF
24
Therefore, from 23 we can write: ).(sin 2 xF
25
Replacing 24 and 25 in 19 we obtain:
.2cos1
32242
42
2 ntdea
txe
n
n
26
Now, if x varies from d to e , then n2cos varies from )1cos( n to
)cos( and consequently n2cos possesses at least 22 n extermal points,
where it assumes alternately the values 1 . Therefore
np2 is the best
approximation out of nP2 , and 14 will be proved with considering 26 .
Theorem 3. The best uniform polynomial approximation out of nP2
for 12 cax where 0
a
c, on ed , will be:
27),(
1
132)(1
1
32
1
162242
22
2
1
0
2
42
2
42
2*
2 xTtdea
txTt
tdea
t
tdea
txp n
nn
k
n
k
kk
n
and 242
42*
22
1
32
tdea
tpffE
n
nn .
28 where
1,241 2
t
a
cde
a
c
det , nxTn cos)(
wherede
x
2cos .
Iranian Journal of Optimization, Vol 7, Issue 1,winter 2015 761
Proof: The proof is similar to the proof of theorem 2.
If we obtain the best uniform polynomial approximation for 225
1
xxf
on
3,3 by using theorem 2, with 3n , xp*
6 we will see that this result is
similar to the best uniform polynomial approximation obtained in [6]. If we obtain
the best uniform polynomial approximation for 225
1
xxf
on 5,5 by
using theorem 3, with 2n , xp*
4 we will see that this result is similar to the best
uniform polynomial approximation obtained in [6].
Example 1. In figure 1, the function 192
12
x
xf has been drawn. The
dashed points show the best uniform polynomial approximation of degree 8
, xp*
8 , to 12 192
x on 2,1 .
Figure 1: The best approximation of 11922
x .
Example 2. In figure 2, both the function and its best uniform polynomial
approximation, xp*
16 , ( the dashed point ) of degree 16 to 125
x on 2,0 has
been shown.
Iranian Journal of Optimization, Vol 7, Issue 1,winter 2015 762
Figure 2: The best approximation of 125
x .
3. Best Approximation of 12 cbxax
In this section, by using the previous theorems, we obtained the best polynomial
approximation for 12 cbxax on 1,1 .
Theorem 4. The best uniform polynomial approximation out of nP2
for 12 cbxax on 1,1 is as follows:
29,
21
8
21
8
1
41)()( 224
22
2
1
0
2
4
2
4
2
2
a
bxT
t
t
a
bxTt
t
t
t
t
axpa n
n
k
n
k
k
n
where,
2 2
2 2
2 2
4 41, 4 4 0 , 1. 30
4 4
b ac b act b ac a t
a a
31,
21
18
21
1
8
1
41)()( 224
22
2
1
0
2
4
2
4
2
2
a
bxT
t
t
a
bxTt
t
t
t
t
axpb n
nn
k
n
k
kk
n
where,
2 2
2
2 2
4 41 , 4 0, 1. 32
4 4
b ac b act b ac t
a a
Proof: We can write the function 12 cbxax in the form of:
Iranian Journal of Optimization, Vol 7, Issue 1,winter 2015 763
33
.
2
222
4
4
2
11
a
acb
a
bxa
cbxax
therefore 1,1xSince
1 ,1 , .2 2 2
b b bx d e
a a a
34
Now, by changing x to a
bx
2 in theorems 2 and 3, we have:
.
22arccos2cos
21
21
22
arccos2cos2
22
a
bxT
abxn
ab
ab
abx
na
bxT nn
Case1 : ( 042 acb ) In this case, replacing a
cby
2
2
4
4
a
acb , according to 34 ,
the defined t in theorem 2, changes to 30 where 22 44 aacb . Therefore, we
can prove a by using theorem 2.
Case2 : ( 042 acb ) In this case, replacing a
cby
2
2
4
4
a
acb , according to 34 ,
the defined t in theorem 3, changes to 32 . Therefore, we can prove b by using
theorem 3.
Example 3. In figure 3, we have drawn the function 152
12
xx
xf . Also,
the dashed points show the best uniform polynomial approximation of degree 6
, xp*
6 , to 12 152
xx on 1,1 .
Iranian Journal of Optimization, Vol 7, Issue 1,winter 2015 764
Figure 3: The best approximation of 11522 xx .
Example 4. In figure 4, both the function and its best uniform polynomial
approximation , xp*
16 , ( the dashed point ) of degree 16 to 12 62
xx
on 1,1 are shown.
Figure 4: The best approximation of 1622 xx .
4. Conclusion
As seen in this article, in the sequel of previous researches, the best uniform
approximation for 12 cax was achieved. In this case, we applied the interval
ed , as general in place of 1,1 .
Iranian Journal of Optimization, Vol 7, Issue 1,winter 2015 765
Also, by characterizing the best uniform approximation for 12 cbxax on
1,1 , a more general form than previous approximation in [6,8] was obtained.
References
[1]-Achieser N. I., Theory of Approximation, Ungar, New York, 1956.
[2]-Achieser N.I., Theory of Approximation, Dover, New York, translated from
Russian, 1992.
[3]-Bernstein S.N., Extremal Properties of Polynomials and the Best
Approximation of Continuous Functions of Single Real Variable, State United
Scientific and Technical Publishing House, translated from Russian, 1937.
[4]-Cheney E.W., Introduction to Approximation Theory, Chelsea, New York,
1982.
[5]-Dehghan M., Eslahchi M.R., Best uniform polynomial approximation of some
rational functions,Computers and Mathematics with applications, 2009.
[6]-Eslahchi M.R., Dehghan M., The best uniform polynomial approximation to
class of the form 122 xa , Nonlinear Anal., TMA 71 (740_750) , 2009.
[7]-Golomb M., Lectures on theory of approximation, Argonne National
Laboratory, Chicago, 1962.
[8]-Jokar S., Mehri B., The best approximation of some rational functions in
uniformnorm, Appl. Numer. Math. 55 (204-214) , 2005.
[9]-Lam B. , Elliott D.,Explicit results for the best uniform rational approximation
to certain continuous functions, J. Approximation Theory 11 (126-133) , 1974.
[10]-Lorentz G.G., Approximation of Functions, Holt, Rinehart and Winston,
New York, 1986.
[11]-Lubinsky D. S., Best approximation and interpolation of 121
ax and
itstransforms, J. Approx. Theory 125 ( 106-115) , 2003.
[12]-Mason J. C., Handscomb D. C., Chebyshev polynomials, Chapman &
Hall/CRC, 2003.
[13]-Newman D.J.,Rivlin T. J., Approximation of monomials by lower degree
polynomials, Aeq. Math. 14 (451-455) , 1976.
[14]-Ollin H. Z., Best polynomial approximation to certain rational functions,
J.Approx. Theory 26 (389-392) , 1979.
[15]-Rivlin T. J., An introduction to the approximation of functions, Dover, New
York, 1981.
Iranian Journal of Optimization, Vol 7, Issue 1,winter 2015 766
[16]-Rivlin T. J., Chebyshev Polynomials. New York: Wiley, 1990.
[17]-Rivlin T. J., Polynomials of best uniform approximation to certain rational
functions,Numer. Math. 4 (345-349) , 1962.
[18]-Timan A.F., Theory of Approximation of a Real Variable, Macmillan, New
York, translated from Russian, 1963.
[19]-Watson G. A., Approximation Theory and Numerical Methods Chicago,
John Wiley & Sons, 1980.