Mathematical Foundations

for Data Science

BITS Pilani
Pilani|Dubai|Goa|Hyderabad
 Orthogonal Matrix
• Let A be a real square matrix such that

AA  A A  I  A1  AT
T T

Then A is called an Orthogonal Matrix

• The determinant of an Orthogonal matrix has
the value +1 or -1.

BITS Pilani, Deemed to be University under Section 3 of UGC Act, 1956

• Examples
1 2 2 cos 0 sin 
A   2 1  2 A   0 0 
1 1
3 
2 2 1  sin  0 cos 

• Orthogonal transformations are

transformations of the form
y  Ax
where A is an orthogonal matrix.

BITS Pilani, Deemed to be University under Section 3 of UGC Act, 1956

• An orthogonal transformation preserves the
value of the inner product of vectors a and b
n
in R .
• Two vectors a1 and a2are said to form an
orthogonal system or pair wise orthogonal if

a a 0 or a a 0
T
1 2
T
2 1

BITS Pilani, Deemed to be University under Section 3 of UGC Act, 1956

• Example:

1 1
Let a1  0 and a2  1
 
1 1 
1
a1  a2  a1 a2  1 0 1 1 
T

1 
 (11)  (01)  (11)  0
BITS Pilani, Deemed to be University under Section 3 of UGC Act, 1956
• A square matrix is orthogonal if and only if its
column vectors a 1 , a 2 , ..........a n
form an orthogonal system.
that is,
aT a  0 for i j
i j
Example:

1 1 1
• A   0 2  1 is an Orthogonal matrix.
 
1 1 1 

BITS Pilani, Deemed to be University under Section 3 of UGC Act, 1956

 Observations
1. Eigen Values of a Symmetric matrix are real.
2. Eigen Values of a Skew-Symmetric matrix are
pure imaginary or zero.
3. Eigen Values of an Orthogonal matrix are real
or complex conjugates in pairs and have
absolute value 1.
4. If an nxn matrix A has n distinct eigen values

then the corresponding n eigen vectors are L.I

BITS Pilani, Deemed to be University under Section 3 of UGC Act, 1956

 Examples
3 1 1 
1. A  1 3 1 eigen values are 1,4,4.
1 1 3 
0 9 12
2. A  9 0 20  eigen values are 0,-25i,25i.
12  20 0 

1 2 2
A  3  2 1  2
1
3. eigen values are -1,
 2 2 1 (5+i√11)/6, (5-i√11)/6.
 

BITS Pilani, Deemed to be University under Section 3 of UGC Act, 1956

 Properties of Eigen Values

• Product of eigen values of a matrix A is equal to

determinant of A and
Sum of eigen values is Trace A.
• If  is an eigen value of a matrix , the is the
1
1 
eigen values of A .
• A square matrix A and its transpose AT have
same eigen values.

• If is an eigen value of a matrix A, then n

is the eigen value of An.

BITS Pilani, Deemed to be University under Section 3 of UGC Act, 1956
 Exercise
1. Find the sum and product of the eigen values of
 2 3  2 7 2 2
(i) A   2 1 1  (ii) A  6 1 2 
1 0 2  6 2 1
3 1
A
1. Find the eigen values of and A where
1 1 3
A  1 5 1
 
3 1 1

BITS Pilani, Deemed to be University under Section 3 of UGC Act, 1956

Similarity of Matrices
An n x n matrix A^ is called similar to an n x n matrix A if
A^=P-1AP
For some nonsingular matrix n x n matrix P

Theorem:
Diagonalization

5 4
A 1 2
Gershgorin’s Theorem
Quadratic Forms and
Transformation to Principal Axis
Let
Q=xT Ax, assume A real symmetric
There A has an orthonormal basisof n eigen vectors
A=XDX-1 = XDXT
set XTx=y Q=yTDy
Transformation to Principal axes,
conic section
Thank you

12

BITS Pilani, Deemed to be University under Section 3 of UGC Act, 1956