The Stone-Weierstrass Theorem

Approximating continuous functions by smooth functions

Satvik Saha
July 31, 2021
Summer Programme
Indian Institute of Science Education and Research, Kolkata
Approximation in metric spaces
Approximation

The object α approximates β, to some degree of accuracy .

d(α, β) < 

α lies within a narrow region centred at β.

1
Approximation

The sequence {αn } converges to β.

Given any , there exists n0 such that for all n ≥ n0 ,

d(αn , β) < 

Every neighbourhood of β contains some tail of {αn }.

2
Approximation

1 1 1 π
1− + − + ··· →
3 5 7 4

1 1 1
x − x3 + x5 − x7 + · · · → arctan x
3 5 7

This series converges uniformly on [−1, +1].

3
Approximating arctan x

4
Integrating a uniformly convergent series

A power series converges uniformly on its interval of

convergence.

∞
x2k
(−1)n
X
cos x =
(2k)!
n=0
∞
x2k+1
(−1)n
X
sin x =
(2k + 1)!
n=0

Z b
cos x dx = sin b − sin a.
a

5
Integrating a uniformly convergent series

If fn → f uniformly on [a, b], then given  > 0, there exists

n0 ∈ N such that for all n ≥ n0 and x ∈ [a, b],

|fn (x) − f (x)| < .

Z b
−(b − a) < fn (x) − f (x) dx < (b − a)
a

Z b Z b
lim fn (x) dx = f (x) dx
n→∞ a a

6
Metric spaces of functions

For every pair of real functions f , g on E, define

d(f , g) = kf − gk∞ = sup |f (x) − g(x)|.

x∈E

• d(f , g) ≥ 0, and d(f , g) = 0 if and only if f = g.

• d(f , g) = d(g, f ).
• d(f , h) ≤ d(f , g) + d(g, h).

We require d(f , g) to be finite, so all functions under

consideration must be bounded.

7
Metric spaces of functions

The function g approximates f , to some degree of accuracy .

d(g, f ) < 

The curve g lies within a narrow strip centred at f .

f − ≤ g ≤ f +

8
Metric spaces of functions

The sequence {fn } converges to f .

Given any , there exists n0 such that for all n ≥ n0 ,

d(fn , f ) < 

The sequence fn → f uniformly.

9
Given a real valued function f on a domain X, can we find a
sequence of ‘nice’ functions (typically polynomials) which
converge uniformly to f ?

10
Uniform Limit Theorem

Let {fn } be a sequence of continuous, real valued functions on

X. If fn → f uniformly on X, then f is continuous.

Fix x0 ∈ X. Then for sufficiently high n, there is a δ

neighbourhood of x0 on which

|f (x) − f (x0 )| ≤ |f (x) − fn (x)| + |fn (x) − fn (x0 )| + |fn (x0 ) − f (x0 )|
< 3.

11
The Weierstrass Approximation Theorem

Given any real valued, continuous function f on a compact

interval [a, b], there exists a sequence of polynomials {pn }
such that pn → f uniformly on [a, b].

The Bernstein polynomials can be used to prove this

constructively.

12
The Bernstein polynomials
Bernstein polynomials

The Bernstein polynomials are defined as

n k
 
k
Bn (x) = x (1 − x)n−k
k

Note that Bkn peaks at x = k/n.

The Bernstein expansion of a function on [0, 1] is defined as

n  
n k k
 
n−k
X
Bn (f , x) = x (1 − x) f
k n
k=0

13
Bernstein polynomials

14
Bernstein polynomials

14
Bernstein polynomials

14
Bernstein polynomials

14
Bernstein polynomials

14
Bernstein polynomials

14
Bernstein polynomials

14
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

15
Bernstein expansion of ex cos(2πx)

16
Generalizing polynomials
Algebras of functions

A collection of real valued functions A on a set E is called an

algebra if

• f ∈ A, g ∈ A =⇒ f + g ∈ A
• f ∈ A, g ∈ A =⇒ fg ∈ A
• f ∈ A, c ∈ R =⇒ cf ∈ A

If f ∈ A and p is a polynomial, then p ◦ f ∈ A.

17
Interpolation

An algebra A vanishes at no point of E if given x ∈ E, there

exists f ∈ A such that f (x) 6= 0.
An algebra A separates points of E if given distinct x1 , x2 ∈ E,
there exists f ∈ A such that f (x1 ) 6= f (x2 ).

Let the algebra A vanish at no point of E and separate points

of E. Given distinct x1 , x2 ∈ E and c1 , c2 ∈ R, there exists f ∈ A
such that f (x1 ) = c1 , f (x2 ) = c2 .

18
Interpolation

Let f1 , f2 ∈ A such that f1 (x1 ) 6= 0 and f2 (x2 ) 6= 0, and let g ∈ A

such that g(x1 ) 6= g(x2 ). Define the functions

g − g(x2 ) f1 g − g(x1 ) f2
h1 = , h2 = .
g(x1 ) − g(x2 ) f1 (x1 ) g(x2 ) − g(x1 ) f2 (x2 )

Note that hi (xj ) = δij . Finally, set

f = c1 h1 + x2 h2 .

Note that this can be extended to arbitrarily many points, in

the manner of Lagrange interpolation.

19
Interpolation and continuity

Let f , g be continuous, real valued functions on X such that

f (x0 ) = g(x0 ) for some x0 ∈ X. Then, g approximates f to an
arbitrary degree of accuracy  on some neighbourhood of x0 .

Note that h = g − f is continuous, hence the pre-image of the

open interval (−, +) is some open set U ⊆ X containing x0 .
Thus, on some neighbourhood Nδ (x0 ) ⊆ U, we have

− < g − f < 

f −<g<f +

20
Closure

The set of uniform limits of functions from an algebra is called

its uniform closure.

A uniformly closed algebra contains all uniform limits of its

functions.

The uniform closure B of an algebra A of bounded functions

is a uniformly closed algebra.

21
Closure

Let A be an algebra of real valued, bounded functions on X,

and let B be its uniform closure. If f ∈ B, then |f | ∈ B.

Let  > 0, let M be such that |f | < M. Pick a polynomial p such

that for all |x| < M,
|p(x) − |x|| < .

Then, for all x ∈ X, we have |f (x)| < M so

|p(f (x)) − |f (x)|| < .

Finally, note that p ◦ f ∈ B.

22
Closure

If f , g ∈ B, then max(f , g) ∈ B and min(f , g) ∈ B.

Note that
1 1
max(f , g) = (f + g) + |f − g|,
2 2
1 1
min(f , g) = (f + g) − |f − g|.
2 2

This gives us a way of ‘stitching’ functions from B together.

23
The Stone-Weierstrass Theorem
The Stone-Weierstrass Theorem

Let K be a compact metric space, and let A be an algebra of

real continuous functions on K which separates points of K
and vanishes at no point of K. The uniform closure of A
consists of all real valued, continuous functions on K.

In other words, given any real valued continuous function f on

K, there exists a sequence of functions {fn } from A such that
fn → f uniformly on K.

24
f −  < gst for all x ∈ Ust

25
f −  < gst for all x ∈ Ust

25
f −  < gst for all x ∈ Ust

25
f −  < gs for all x ∈ K

25
f −  < gs < f +  for all x ∈ Us

26
f −  < gs < f +  for all x ∈ Us

26
f −  < gs < f +  for all x ∈ Us

26
f −<g<f + for all x ∈ K

26
Polynomials

Let P be the set of polynomials on Rn , for some n ≥ 1. Then,

P is an algebra of real continuous functions which separates
points on Rn and vanishes at no point of Rn .
In a sense, P is the algebra generated by the projection maps,
πi (x1 , . . . , xn ) = xi , along with the constant functions.

Any real continuous function on Rn can be uniformly

approximated on a compact subset of Rn .

27
Fourier-like functions

Let F be the set of functions of the form

N
f : S1 → R, eix 7→ a0 +
X
an cos(nx) + bn sin(nx).
n=1

Here, we identify S1 with the unit circle in C; S1 is compact.

Also, F is an algebra of real continuous functions which
separates points on S1 and vanishes at no point of S1 .

Any real continuous function on S1 can be uniformly

approximated by functions from F.

28