Lecture Static 04 - 014 PDF

Chebyshev Polynomials
Least Squares, redux
Outline
Numerical Analysis and Computing

Lecture Notes #12
Approximation Theory
Chebyshev Polynomials & Least Squares, redux
Orthogonal Polynomials
Chebyshev Polynomials, Intro & Definitions
Properties

Examples
More than one variable? No problem!
Joe Mahaffy,
hmahaffy@math.sdsu.edui
Department of Mathematics
Dynamical Systems Group

Computational Sciences Research Center
San Diego State University

San Diego, CA 92182-7720
http://www-rohan.sdsu.edu/jmahaffy
Spring 2010
Joe Mahaffy, hmahaffy@math.sdsu.edui
(1/45)
Properties
Orthogonal Polynomials: A Quick Summary
The ideas and techniques we developed i.e. Gram-Schmidt

orthogonalization with respect to a weight function over any
interval have applications far beyond least squares problems.
(2/45)
Properties
Background
The Legendre polynomials are solutions to the Legendre

Differential Equation (which arises in numerous problems
exhibiting spherical symmetry)
(1 x 2 )
d 2y
dy
2x
+ ( + 1)y = 0,
2
dx
dx
or equivalently
The Legendre Polynomials are orthogonal on the interval [1, 1]

with respect to the weight function w (x) = 1. One curious
property of the Legendre polynomials is that their roots (all real)
yield the optimal node placement for Gaussian quadrature.
The Legendre Polynomials
So far we have seen the use of orthogonal polynomials can help us

solve the normal equations which arise in discrete and continuous
least squares problems, without the need for expensive and
numerically difficult matrix inversions.
(3/45)
d
2 dy
(1 x )
+ ( + 1)y = 0,
dx
dx
Applications: Celestial Mechanics (Legendres original application), Electrodynamics, etc...

(4/45)
Properties
Other Orthogonal Polynomials
Background
Orthogonal polynomials have very useful properties in the solution

of mathematical and physical problems. [... They] provide a
natural way to solve, expand, and interpret solutions to many types
of important differential equations. Orthogonal polynomials are
especially easy to generate using Gram-Schmidt
orthonormalization.
The roots of orthogonal polynomials possess many rather
surprising and useful properties.
Properties
The Laguerre Polynomials
Background
The Laguerre polynomials are solutions to the Laguerre

differential equation
x
(http://mathworld.wolfram.com/OrthogonalPolynomials.html)
d2
dy
+ (1 x)
+ y = 0
2
dx
dx
They are associated with the radial solution to the Schrodinger

equation for the Hydrogen atoms electron (Spherical Harmonics).

Properties
More Orthogonal Polynomials

Polynomials
Chebyshev (1st)
Chebyshev (2nd)
Gegenbauer
Hermite
Jacobi
Legendre
Laguerre
Laguerre (assoc)
(5/45)
Background
Interval
[1, 1]
[1, 1]
[1, 1]
(, )
(1, 1)
[1, 1]
[0, )
[0, )
w(x)
2
1/
1x
1 x2
(1 x 2 )1/2
2
e x
(1 x) (1 + x)
1
e x
x k e x
(6/45)
Properties
1 t 2 t (n+1)
dt
(1 2tz + t 2 )
Chebyshev Polynomials are used to minimize approximation

error. We will use them to solve the following problems:
These are the Hermite orthogonal polynomials, not to be confused

with the Hermite interpolating polynomials...
Chebyshev Polynomials: The Sales Pitch.
1
Tn (z) =
2i
Today well take a closer look at Chebyshev polynomials of the first

kind.
(7/45)
[1] Find an optimal placement of the interpolating points

{x0 , x1 , . . . , xn } to minimize the error in Lagrange interpolation.
[2] Find a means of reducing the degree of an approximating polynomial with minimal loss of accuracy.
(8/45)
Properties
Chebyshev Polynomials: Definitions.
Chebyshev Polynomials, Tn (x), n 2.
The Chebyshev polynomials {Tn (x)} are orthogonal onthe interval

(1, 1) with respect to the weight function w (x) = 1/ 1 x 2 , i.e.
Z
hTi (x), Tj (x)iw (x)
Ti (x)Tj (x) w (x)dx = i i,j .
We introduce the notation = arccos x, and get

Tn ((x)) Tn () = cos(n),
Definition (Chebyshev Polynomials)
Tn (x) = cos(n arccos x),
Tn+1 () = cos((n + 1)) = cos(n) cos() sin(n) sin()

Tn1 () = cos((n 1)) = cos(n) cos() + sin(n) sin()
Tn+1 () + Tn1 () = 2 cos(n) cos().
Returning to the original variable x, we have
For x [1, 1], define
T0 (x) = cos(0) = 1,
Tn+1 (x) = 2x cos(n arccos x) Tn1 (x),
n 0.
or
Tn+1 (x) = 2xTn (x) Tn1 (x).
T1 (x) = x.
(9/45)
Properties
where [0, ].
We can find a recurrence relation, using these observations:
We could use the Gram-Schmidt orthogonalization process to find

them, but it is easier to give the definition and then check the
properties...
Note:
Properties
The Chebyshev Polynomials
Chebyshev Polynomials & Least Squares, redux (10/45)

Properties
Orthogonality of the Chebyshev Polynomials, I
0.5
T1(x)
T2(x)
T3(x)
T4(x)
T5(x)
Tn (x)Tm (x)
dx =
1 x2
cos(n arccos x) cos(m arccos x)
dx
.
1 x2
Reintroducing = arccos x gives,

d =
dx
,
1 x2
and the integral becomes

Z
Z 0
cos(n) cos(m) d =
-0.5
cos(n) cos(m) d.
Now, we use the fact that

-1
-1
-0.5
0.5
cos(n) cos(m) =
cos(n + m) + cos(n m)
...
2
Properties
Orthogonality of the Chebyshev Polynomials, II

We have:
If m 6= n, we get
1
1
sin((n + m)) +
sin((n m)) = 0,
2(n + m)
2(n m)
0
x
1
sin((n + m)) +
2(n + m)
2
.
2

Properties
Zeros and Extrema of Chebyshev Polynomials Proof.

Proof.
Let:
xk = cos
2k 1
,
2n
xk = cos
k
n
=
=
Tn (xk )
Tn (xk )
n sin(k)
sin( k
n )
= 0,
Properties
n (x) by dividing Tn (x)

We get the monic Chebyshev polynomials T
n1
by 2 , n 1. Hence,
0 (x) = 1,
T
n (x) =
T
1
2n1
Tn (x),
n 1.
They satisfy the following recurrence relations

= cos(k) = (1)k .
cos n arccos cos k
n
A monic polynomial is a polynomial with leading coefficient 1.
n sin(n arccos(x))
d
,
[cos(n arccos(x))] =
dx
1 x2
n sin(n arccos(cos( k
n )))
q
1cos2 ( k
)
n
Definition (Monic Polynomial)
cos(n arccos(xk )) = cos n arccos cos 2k1
2n
cos 2k1
= 0,
2
Tn (x) =
Moreover, Tn (x) assumes its absolute extrema at

k
, with Tn (xk ) = (1)k , k = 1, . . . , n 1.

xk = cos
n
Monic Chebyshev Polynomials, I
Then:
Tn (xk )
Theorem
The Chebyshev polynomial of degree n 1 has n simple zeros in
[1, 1] at
2k 1
xk = cos
, k = 1, . . . , n.
2n
Payoff: No matter what the degree of the polynomial, the oscillations are kept under control!!!
Hence, the Chebyshev polynomials are orthogonal.

Properties
Zeros and Extrema of Chebyshev Polynomials.
cos(n + m) + cos(n m)
d.
2
if m = n, we have
2 (x) = x T
1 (x) 1 T
T
2 0 (x)
n+1 (x) = x T
n (x) 1 T
T
4 n1 (x).
Properties
Monic Chebyshev Polynomials, II
The Monic Chebyshev Polynomials
n (x) coincides with

The location of the zeros and extrema of T
those of Tn (x), however the extreme values are
k
n (xk ) = (1) ,
T
2n1
Properties
k = 1, . . . , n 1.
0.5
Definition
Let Pn denote the set of all monic polynomials of degree n.
T1(x)
T2(x)
T3(x)
T4(x)
T5(x)
Theorem (Min-Max)
The monic Chebyshev polynomials Tn (x), have the property that
1
n (x)| max |Pn (x)|,
= max |T
2n1
x[1,1]
x[1,1]
Pn (x) Pn .
n (x).
Moreover, equality can only occur if Pn (x) T
-1
-1
If x0 , x1 , . . . , xn are distinct points in the interval [1, 1] and

f C n+1 [1, 1], and P(x) the nth degree interpolating Lagrange
polynomial, then x [1, 1] (x) (1, 1) so that
n
f (n+1) ((x)) Y
f (x) P(x) =
(x xk ).
(n + 1)!
k=0
We have no control over f (n+1) ((x)), but we can

Qnplace the nodes
in a clever
Q way as to minimize the maximum of k=0 (x xk ).
Since nk=0 (x xk ) is a monic polynomial of degree (n + 1), we
know the min-max is obtained when the nodes are chosen so that
n
Y
2k + 1
(x xk ) = Tn+1 (x), i.e. xk = cos

.
2(n + 1)
k=0
-0.5
Properties
Optimal Node Placement in Lagrange Interpolation, I
-0.5
0.5

Properties
Optimal Node Placement in Lagrange Interpolation, II

Theorem
If P(x) is the interpolating polynomial of degree at most n with
nodes at the roots of Tn+1 (x), then
max |f (x) P(x)|
x[1,1]
2n (n
1
max |f (n+1) (x)|,
+ 1)! x[1,1]
f C n+1 [1, 1].

Extending to any interval: The transformation
x =
1
[(b a)x + (a + b)]
2
transforms the nodes xk in [1, 1] into the corresponding nodes xk

in [a, b].
Properties
Example: Interpolating f (x) = x 2 e x .
Properties
Example: Interpolating f (x) = x 2 e x The Error.
3
f(x)
Taylor2(x)
Lagrange2EquiSpaced(x)
Lagrange2Optimal(x)
2.5
Deviation: EquiSpaced
Deviation: Optimal
0.1
1.5
0
0.5
-0.1
0
0
0.2
0.4

0.8
0.6

Properties
0.01
0.005
0.05
0.1
0.005
0.15
0.01
1.5

Examples
Introduction
Error: EquiSpaced
Error: Optimal
0.015
0.2
0.8
Degree 3
0.05
0.4
0.6
0.8
Before we move on to new and exciting orthogonal polynomials

with exotic names... Lets take a moment (or two) and look at the
usage of Least Squares Approximation.
0.015
0
0.2
0.4
10
0.6
0.8
Degree 5
Degree 4
0.6
0.02
Error: EquiSpaced
Error: Optimal
x 10
Least Squares: Revisited
Degree 2
0.15
0.2
0
0.4
Example: Interpolating f (x) = x 2 e x The Error.
0.1
0.2
x 10
This section is a how-to with quite a few applied example of

least squares approximation...
Error: EquiSpaced
Error: Optimal
Error: EquiSpaced
Error: Optimal
1
0.5
0.5
1.5
0
0.2
0.4
0.6
0.8
5
0
0.2
0.4
0.6
0.8
Examples
Example #1 Warm-up
1 of 3
Examples
Example #1 Warm-up
2 of 3
We now have the coefficients for the polynomials, lets plot:

matlab>> xv = 1.0:0.01:2.1;
First we consider the problem of fitting 1st, 2nd, and 3rd degree
polynomials to the following data:
p1 = polyval(flipud(pcoef1),xv);
plot(xv,p3,k-,linewidth,3); hold on;
plot(x,y,ko,linewidth,3); hold off
x = [1.0 1.1 1.3 1.5 1.9 2.1]

y = [1.84 1.90 2.31 2.65 2.74 3.18]
matlab [First we define the matrices]
A1 = [ones(size(x)) x];
A2 = [A1 x.*x];
A3 = [A2 x.*x.*x ];
[Then we solve the Normal Equations]
pcoef1 = A1\y;
pcoef2 = A2\y;
pcoef3 = A3\y;
Optimal Linear Approximation
Optimal Quadratic Approximation
3.2
Note: The matrices A1, A2, and A3 are tall and skinny. Normally we
would compute (An An)1 (An y ), however when matlab encounters
An\y it automatically gives us a solution in the least squares sense.
Optimal Cubic Approximation
3.2
3.2
2.8
2.8
2.8
2.6
2.6
2.6
2.4
2.4
2.4
2.2
2.2
2.2
LSQerror: 0.087656
1.8
LSQerror: 0.069900
1.8
1
1.2
1.4
1.6
1.8
LSQerror: 0.044729
1.8
1
1.2
1.4
1.6
1.8
1.2
1.4
1.6
1.8
Figure: The least squares polynomials p1 (x), p2 (x), and p3 (x).

Examples
Example #1 Warm-up
3 of 3

Examples
Example #2 Something More Exotic
1 of 2
Consider the same data:

Finally, we compute the error
matlab>> p1err = polyval(flipud(pcoef1),x)
x = [1.0 1.1 1.3 1.5 1.9 2.1]

- y;
p2err = polyval(flipud(pcoef2),x) - y;
p3err = polyval(flipud(pcoef3),x) - y;
disp([sum(p1err.*p1err) sum(p2err.*p2err)
sum(p3err.*p3err)])
But lets find the best fit of the form a + b x to this data! Notice
that this expression is linear in its parameters a, b, so we can solve
the corresponding least squares problem!
matlab>>
Which gives us the fitting errors

P1Err = 0.0877,
y = [1.84 1.90 2.31 2.65 2.74 3.18]
A = [ones(size(x)) sqrt(x)];
pcoef = A\y;
P2Err = 0.0699,
P3Err = 0.0447
xv = 1.0:0.01:2.1;
fv = pcoef(1) + pcoef(2)*sqrt(xv);
plot(xv,fv,k-,linewidth,3); hold on;
plot(x,y,ko,linewidth,3); hold off;
Examples
Example #2 Something More Exotic

0.5
Optimal a+b x
2 of 2
Examples
Getting Even More Exotic?
1 of 2
Approximation
3.2
2.8
As long as the model is linear in its parameters, we can solve the

least squares problem.
2.6
Non-linear dependence will have to wait until Math 693a.
We can fit this model:
2.4
M1 (a, b, c, d) = a + bx 3/2 + c/ x + de sin(x)
2.2
2
LSQerror: 0.074920
1.8
1
1.2
1.4
1.6
1.8
Just define the matrix
Figure: The best fit of the form a + b x.
We compute the fitting error:

matlab>> ferr = pcoef(1) + pcoef(2)*sqrt(x)
matlab>>
and compute the coefficients
- y;
matlab>>
disp(sum(ferr.*ferr))

coef = A\y;
etc...
Err
P{a+b
x} = 0.0749
Which gives us
A = [ones(size(x)) x. (3/2) 1./sqrt(x) exp(sin(x))];

Examples
Getting Even More Exotic?

2 of 2

Examples
Getting Multi-Dimensional
Optimal "Exotic" Model

3.2
3
It seems quite unlikely the model
M1 (a, b, c, d) = a + bx 3/2 + c/ x + de sin(x)
2.8
2.6
will ever be useful.
2.4
However, we have forgotten about one important aspect of the

problem so far our models have depended on only one variable,
x.
2.2
2
LSQerror: 0.062771
1.8
How do we go about fitting multi-dimensional data?

1
1.2
1.4
1.6
1.8
The optimal approximation of this form is

11.0059
16.4133 0.9970x 3/2
1.1332e sin(x)
x
Examples
Example: Going 2D
1 of 10
Examples
Example: Going 2D
2 of 10
Lets try to fit a simple 3-parameter model to this data
matlab>> x=1:0.25:5;
y=1:0.25:5;
[X,Y]=meshgrid(x,y);
Fxy=1+sqrt(X)+Y. 3+0.05*randn(size(X));
surf(x,y,Fxy)
M(a, b, c) = a + bx + cy
matlab>>
150
100
50
sz
= size(X);
Bm
= reshape(X,prod(sz),1);
Cm
= reshape(Y,prod(sz),1);
Am
= ones(size(Bm));
RHS
= reshape(Fxy,prod(sz),1);
= [Am Bm Cm];
coef = A \ RHS;
0
6
5
fit
4
3
2
0
surf(x,y,fitError)
Figure: 2D-data set, the vertexes on the surface are our data points.
= coef(1) + coef(2)*X + coef(3)*Y;
fitError = Fxy - fit;
Examples
Example: Going 2D
3 of 10
Fit (Model 1)

Examples
Example: Going 2D
4 of 10
Lets try to fit a simple 4-parameter (bi-linear) model to this data
Error (Model 1)
M(a, b, c) = a + bx + cy + dxy
150
30
matlab>>
sz
= size(X);
10
Bm
Cm
10
Dm
= reshape(X.*Y,prod(sz),1);
Am
= ones(size(Bm));
RHS
= reshape(Fxy,prod(sz),1);
= [Am Bm Cm Dm];
20
100
50
0
6
20
6
5
4
3
2
0
4
3
2
0
Figure: The optimal model fit, and the fitting error for the
least squares best-fit in the model space M(a, b, c) = a +
bx + cy . Here, the total LSQ-error is 42,282.
coef = A \ RHS;
fit
= coef(1) + coef(2)*X + coef(3)*Y + coef(4)*X.*Y;

surf(x,y,fitError)
Examples
Example: Going 2D
5 of 10
Fit (Model 2)
Example: Going 2D
M(a, b, c) = a + bx + cy + dy 2
30
20
100
matlab>>
10
0
50
10
0
6
20
6
5
4
3
4
3
2
0
2
0
Figure: The fitting error for the least squares best-fit in the
model space M(a, b, c) = a + bx + cy + dxy . Still a pretty
bad fit. Here, the total LSQ-error is still 42,282.
7 of 10
Error (Model 3)
Fit (Model 3)
2
0
2
5
3
2
0
4
3
2
0
model space M(a, b, c) = a + bx + cy + dy 2 . We see a
significant drop in the error (one order of magnitude); and
the total LSQ-error has dropped to 578.8.
7 12 of 10
Example: Going 2D
Why not add both the xy and y 2 always?
4
6
0
6
Examples
The change in the LSQ-error as a function of an added term is one

way to decide what is a useful addition to the model.
100
50
We notice something interesting: the addition of the xy -term to

the model did not produce a drop in the LSQ-error. However, the
y 2 allowed us to capture a lot more of the action.

Example: Going 2D
sz
= size(X);
Bm
Cm
Dm
= reshape(Y.*Y,prod(sz),1);
Am
= ones(size(Bm));
RHS = reshape(Fxy,prod(sz),1);
A
= [Am Bm Cm Dm];
coef = A \ RHS;
fit = coef(1) + coef(2)*X + coef(3)*Y + coef(4)*Y.*Y;
surf(x,y,fitError)
Examples
150
6 of 10
Since the main problem is in the y -direction, we fit try a

4-parameter model with a quadratic term in y
Error (Model 2)
150
Examples
(A)
(AT A)
xy
86.2
7,422
y2
107.3
11,515
Both
170.5
29,066
Table: Condition numbers for the A-matrices (and

associated Normal Equations) for the different models.
Examples
Example: Going 2D
8 of 10
Examples
Example: Going 2D
We fit a 5-parameter model with a quadratic term in y
9 of 10
Fit (Model 4)
Error (Model 4)
M(a, b, c) = a + bx + cy + dy 2 + ey 3
150
matlab>>
sz
Bm
Cm
Dm
Em
Am
RHS
A
coef
fit
=
=
=
=
=
=
=
=
=
=
size(X);
reshape(X,prod(sz),1);
reshape(Y,prod(sz),1);
reshape(Y.*Y,prod(sz),1);
reshape(Y.*Y.*Y,prod(sz),1);
ones(size(Bm));
reshape(Fxy,prod(sz),1);
[Am Bm Cm Dm Em];
A \ RHS;
coef(1) + coef(2)*X + coef(3)*Y + coef(4)*Y.*Y +
50
0.1
0
6
a + bx
a + bx
a + bx
a + bx
a + bx
+ cy
+ cy
+ cy
+ cy
+ cy
+ dxy
+ dy 2
+ ey 3
+ dy 2 + ey 3
(AT A)
42,282
42,282
578.8
2.695
0.9864
278
7,422
11,515
107,204
1,873,124
4
3
2
0

Table: Summary of LSQ-error and conditioning of the Normal

Equations for the various models. We notice that additional
columns in the A-matrix (additional model parameters) have a
severe effect on the conditioning of the Normal Equations.
model space M(a, b, c) = a + bx + cy + dy 2 + ey 3 . We
now have a pretty good fit. The LSQ-error is now down to
0.9864.
Examples
LSQ-error
2
0
10 of 10
Model
4
3
Example: Going 2D
0.2
6

surf(x,y,fitError)
0.1
100
coef(5)*Y. 3;
0.2

Examples
Moving to Even Higher Dimensions
At this point we can state the Linear Least Squares fitting problem
in any number of dimensions, and we can use exotic models if we
want to.
In 3D we need 10 parameters to fit a model with all linear, and
second order terms
M(a, b, c, d, e, f , g , h, i, j) =
a + bx + cy + dz + ex 2 + fy 2 + gz 2 + hxy + ixz + jyz
With nx , ny , and nz data points in the x-, y -, and z-directions
(respectively) we end up with a matrix A of dimension
(nx ny nz ) 10.
Examples
Ill-conditioning of the Normal Equations
Needless(?) to say, the normal equations can be quite

ill-conditioned in this case. The ill-conditioning can be eased by
searching for a set of orthogonal functions with respect to the
inner product
Z
hf (x), g (x)i =
xb
xa
yb
ya
zb
za
f (x, y , z)g (x, y , z) dx dy dz
Thats *sometimes* possible, but well leave the details as an

exercise for a dark and stormy night...

Lecture Static 04 - 014 PDF

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

Lecture Static 04 - 014 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lecture Static 04 - 014 PDF

Uploaded by

Copyright:

Available Formats

Chebyshev Polynomials

Least Squares, redux

Numerical Analysis and Computing

Least Squares, redux

Dynamical Systems Group

San Diego State University

Chebyshev Polynomials & Least Squares, redux

Orthogonal Polynomials: A Quick Summary

The ideas and techniques we developed i.e. Gram-Schmidt

The Legendre polynomials are solutions to the Legendre

The Legendre Polynomials are orthogonal on the interval [1, 1]

Chebyshev Polynomials & Least Squares, redux

The Legendre Polynomials

So far we have seen the use of orthogonal polynomials can help us

Joe Mahaffy, hmahaffy@math.sdsu.edui

Joe Mahaffy, hmahaffy@math.sdsu.edui

Applications: Celestial Mechanics (Legendres original application), Electrodynamics, etc...

Chebyshev Polynomials & Least Squares, redux

Other Orthogonal Polynomials

Orthogonal polynomials have very useful properties in the solution

The Laguerre Polynomials

The Laguerre polynomials are solutions to the Laguerre

They are associated with the radial solution to the Schrodinger

Chebyshev Polynomials & Least Squares, redux

More Orthogonal Polynomials

Chebyshev Polynomials & Least Squares, redux

Chebyshev Polynomials are used to minimize approximation

These are the Hermite orthogonal polynomials, not to be confused

Chebyshev Polynomials & Least Squares, redux

Chebyshev Polynomials: The Sales Pitch.

Today well take a closer look at Chebyshev polynomials of the first

Joe Mahaffy, hmahaffy@math.sdsu.edui

[1] Find an optimal placement of the interpolating points

Chebyshev Polynomials & Least Squares, redux

Chebyshev Polynomials: Definitions.

Chebyshev Polynomials, Tn (x), n 2.

The Chebyshev polynomials {Tn (x)} are orthogonal onthe interval

hTi (x), Tj (x)iw (x)

Ti (x)Tj (x) w (x)dx = i i,j .

We introduce the notation = arccos x, and get

Definition (Chebyshev Polynomials)

Tn (x) = cos(n arccos x),

Joe Mahaffy, hmahaffy@math.sdsu.edui

Tn+1 () = cos((n + 1)) = cos(n) cos() sin(n) sin()

For x [1, 1], define

Tn+1 (x) = 2x cos(n arccos x) Tn1 (x),

Tn+1 (x) = 2xTn (x) Tn1 (x).

Chebyshev Polynomials & Least Squares, redux

Joe Mahaffy, hmahaffy@math.sdsu.edui

We can find a recurrence relation, using these observations:

We could use the Gram-Schmidt orthogonalization process to find

The Chebyshev Polynomials

Chebyshev Polynomials & Least Squares, redux (10/45)

Orthogonality of the Chebyshev Polynomials, I

cos(n arccos x) cos(m arccos x)

Reintroducing = arccos x gives,

and the integral becomes

Now, we use the fact that