Scribd
Upload
38 views

Approximation PDF

Uploaded by

Deepak Malik
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
38 views

Approximation PDF

Uploaded by

Deepak Malik
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

A Probabilistic Proof of the Weierstrass Approximation Theorem

Kenneth M. Levasseur

The American Mathematical Monthly, Vol. 91, No. 4. (Apr., 1984), pp. 249-250.

Stable URL:
http://links.jstor.org/sici?sici=0002-9890%28198404%2991%3A4%3C249%3AAPPOTW%3E2.0.CO%3B2-K

The American Mathematical Monthly is currently published by Mathematical Association of America.

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained
prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in
the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/journals/maa.html.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic
journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers,
and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take
advantage of advances in technology. For more information regarding JSTOR, please contact [email protected].

http://www.jstor.org
Thu Mar 6 22:03:45 2008
NOTES

EDITEDBY SABRAS. ANDERSON,


SHELDONAXLER,AND J. ARTHURSEEBACH,
JR.

For instructions about submitting Notes for publication in this department see the inside front cover.

A PROBABILISTICPROOF OF THE WEIERSTRASS APPROXIMATION THEOREM

KENNETH M. LEVASSEUR
Department of Mathematics, University of Lowell, Lowell, MA 01854

Among the proofs of the Weierstrass Approximation Theorem, Bernstein's proof is surely the
most common. It is particularly appealing because it is a constructive proof. Iff is a continuous
real-valued function of the interval [0, 11, the sequence of Bernstein polynomials

converges uniformly to f . This fact establishes


THEWEIERSTRASS APPROXIMATION THEOREM. Iff is a continuous real-valued function on the
interval [O,1] and E > 0, then there exists a polynomialfunction, p , such that maxl f (x) - p(x)l < E
for0 6 x 6 1.
Once you've seen this sequence of polynomials, the actual convergence proof is much less
inspiring. However the use of some elementary probability theory can make the convergence proof
a bit more interesting. This is not an original approach; it is actually a return to Bernstein's
method of proof in [I]. Since this reference is not readily available, it is worth repeating.
The only tools that are needed are the binomial distribution and Chebyshev's Theorem. The
binomial distribution is

where n is the number of trials, x is the probability of success for any one trial, and j is the
number of successes. Chebyshev's Theorem states that if a probability distribution has mean p and
standard deviation a, the probability of obtaining a value that deviates from the mean by at least
k standard deviations is at most l/k2; i.e., Pr(lx - pI > k a ) 6 l/k2.
To prove convergence of the Bernstein polynomials, define a sequence of families of random
variables F,, ,, 0 6 x 6 1 and n = 0,1,2,. . . in terms of a binomial random variable with n trials
and probability of success x. If j successes occur, define the value of F,, to be f (j/n). Then the
expected value of Fn, is

We now prove that E ( 4 , ,) converges uniformly to f on [O,1] as n -+ co.


Let E > 0 be arbitrary. Since f is continuous, it is bounded on [0, 11; therefore there exists M
such that If(x)l 6 M for 0 6 x 6 1, which implies that I f ( x ) - f(y)l G 2 M for 0 G x G 1. The
continuity of f on [0,1] also implies uniform continuity, i.e., there exists 6 > 0 such that if
Ix - yl < 6, then I f ( x ) - f(y)l < 4 2 . Now select a positive integer k such that 2M/k2 < e/2
and a positive integer N such that k / 2 n < 6. Then if n > N and 0 6 x 6 1,
GILBERT STRANG

The first summand is less than ~ / 2 ,by our choice of 6. To bound the second summand, we use
the fact that the mean and standard deviation of b(n, x; k) are nx and ,-/ respectively.
So if

then I j - nxl > k/- and application of Chebyshev's Theorem yields


2M Z b ( n , x ; j ) 6 2M/k2 < e/2.
li-x12k/2fi

This establishes that I f ( x ) - Bn(f ; x)l < E for all x E [O,l] and n l;; N.
Reference
1. S. N. Bernstein, Demonstration du theoreme de Weierstrass fondee sur le calcul des probabilites, Comm. Soc.
Math. Kharkov, 13 (1912) 1-2.

DUALITY IN THE CLASSROOM

GILBERT STRANG
Depurtment of Mathematics, Massachusetts Institute of Technologl~,Cuml~ridge,MA 02139

This note computes the smallest solution to an underdetermined system Ax = b. It is a


problem that puts to use five or six different topics in linear algebra, and at the same time it
illustrates one new idea: the magic of duality. Perhaps the simplest example of duality, in which
we can see directly that a minimum equals a maximum, is the distance between a point and a line.
The minimum distance from the origin to the points on a line (in R ~ equals ) the maximum
distance to the planes through that line. That is evident to the intuition, which is extremely good
at projection, and it is the special case of our problem when the matrix A is two by three.
In general A will be an m by n matrix with m < n; there are more unknowns than equations.
We assume that the rows are independent, so the rank is m, AA= is invertible, and Ax = b has
infinitely many solutions. The problem is to minimize the Euclidean length 1x1 = (x: + . . .
+ and it can be approached in at least four ways:
(i) by introducing Lagrange multipliers and then minimizing quadratics;
(ii) by developing the duality that was intuitive above;
(iii) as an instance of the pseudo-inverse A+ (in this case a right inverse);
(iv) through projection.

You might also like