EIGEN VALUE DECOMPOSITION

Find Eigen Value Decomposition for the matrix

8 −6 2
[−6 7 −4]
2 −4 3
Sol: The characteristic equation is given by the formula |𝐴 − 𝜆𝐼| = 0
8 − 𝜆 −6 2 1 0 0
| −6 7 − 𝜆 −4 | = 0 𝜆𝐼 = 𝜆 [0 1 0]
2 −4 3 − 𝜆 0 0 1
𝜆3 − (∑ 𝑑)𝜆2 − (∑ 𝑚𝑑 )𝜆 − |𝐴| = 0 𝜆 0 0
= [0 𝜆 0]
∑ 𝑑 = 8 + 7 + 3 = 18 (Sum of all diagonal elements)
0 0 𝜆
7 −4 8 2 8 −6
∑ 𝑚𝑑 = | |+| |+| |
−4 3 2 3 −6 7
= (21 − 16) + (24 − 4) + (56 − 36)
= 45 (Summation other than diagonal elements)
8 −6 2
|𝐴| = |−6 7 −4| = 8(21 − 16) + 6(−18 + 8) + 2(24 − 14)
2 −4 3
= 8(5) + 6(−10) + 2(10)
= 40 − 60 + 20 = 60 − 60 = 0
The equation can be written as
𝜆3 − 18𝜆2 + 45𝜆 − 0 = 0
𝜆3 − 18𝜆2 + 45𝜆 = 0
𝜆(𝜆2 − 18𝜆 + 45) = 0
𝜆 = 0 𝑜𝑟 𝜆2 − 18𝜆 + 45 = 0
𝜆2 − 15𝜆 − 3𝜆 + 45 = 0
𝜆(𝜆 − 15) − 3(𝜆 − 15) = 0
(𝜆 − 15)(𝜆 − 3) = 0
⟹ 𝜆 = 15 ⟹ 𝜆 = 3
∴ 𝑇ℎ𝑒 𝑒𝑖𝑔𝑒𝑛 𝑣𝑎𝑙𝑢𝑒𝑠 𝑎𝑟𝑒 𝜆1 = 0, 𝜆2 = 3, 𝜆3 = 15
Now we form the system of equations
[𝐴 − 𝜆𝐼][𝑋] = 0
8 − 𝜆 −6 2 𝑥
[ −6 7 − 𝜆 −4 ] [𝑦] = 0
2 −4 3 − 𝜆 𝑧
(8 − 𝜆)𝑥 − 6𝑦 + 2𝑧 = 0
−6𝑥 + (7 − 𝜆)𝑦 − 4𝑧 = 0 1
2𝑥 − 4𝑦 + (3 − 𝜆)𝑧 = 0
Case 1: When 𝜆 = 0 eq 1 becomes
8𝑥 − 6𝑦 + 2𝑧 = 0 i
−6𝑥 + 7𝑦 − 4𝑧 = 0 ii
2𝑥 − 4𝑦 + 3𝑧 = 0 iii
Applying Cross Multiplication Rule on i and ii
𝑥 𝑦 𝑧
=− =
−6 2 8 2 8 −6
| | | | | |
7 −4 −6 −4 −6 7
𝑥 𝑦 𝑧
⟹ =− =
24 − 14 −32 + 12 56 − 36
𝑥 −𝑦 𝑧
⟹ = = × 10
10 −20 20
𝑥 𝑦 𝑧
⟹ = =
1 2 2
1
At Eigen value 𝜆1 = 0, the eigen vector 𝑋1 = [2]
2
Case 2: When 𝜆 = 3 eq 1 becomes
5𝑥 − 6𝑦 + 2𝑧 = 0 i
−6𝑥 + 4𝑦 − 4𝑧 = 0 ii
2𝑥 − 4𝑦 + 0𝑧 = 0 iii
Applying Cross Multiplication Rule on i and ii
𝑥 𝑦 𝑧
=− =
−6 2 5 2 5 −6
| | | | | |
4 −4 −6 −4 −6 7
𝑥 𝑦 𝑧
⟹ =− =
24 − 8 −20 + 12 20 − 36
𝑥 −𝑦 𝑧
⟹ = = ×8
16 −8 −16
𝑥 𝑦 𝑧
⟹ = =
2 1 −2
2
At Eigen value 𝜆2 = 3, the eigen vector 𝑋2 = [ 1 ]
−2
Case 2: When 𝜆 = 15 eq 1 becomes
−7𝑥 − 6𝑦 + 2𝑧 = 0 i
−6𝑥 − 8𝑦 − 4𝑧 = 0 ii
2𝑥 − 4𝑦 − 12𝑧 = 0 iii
Applying Cross Multiplication Rule on i and ii
 𝑥 𝑦 𝑧
=− =
−6 2 −7 2 −7 −6
| | | | | |
−8 −4 −6 −4 −6 −8
𝑥 𝑦 𝑧
⟹ =− =
24 + 16 28 + 12 56 − 36
𝑥 −𝑦 𝑧
⟹ = = × 20
40 40 20
𝑥 𝑦 𝑧
⟹ = =
2 −2 1
2
At Eigen value 𝜆3 = 15, the eigen vector 𝑋2 = [−2]
1
The eigen values are 𝜆 = 0, 3, 15
1 2 2
Eigen Vectors are 𝑋1 = [2] 𝑋2 = [ 1 ] 𝑋3 = [−2]
2 −2 1
The Eigen Value Decomposition is given by the formula
𝐴 = 𝑃𝐷𝑃 −1
For a symmetric matrix (where row elements are equal to column elements), we
have the formula 𝐴 = 𝑃𝐷𝑃𝑇
where:
𝑃 = 𝐼𝑛𝑖𝑡𝑖𝑎𝑙 𝑀𝑎𝑡𝑟𝑖𝑥
𝐷 = 𝐷𝑖𝑎𝑔𝑜𝑛𝑎𝑙𝑖𝑧𝑒𝑑 𝐸𝑖𝑔𝑒𝑛 𝑉𝑎𝑙𝑢𝑒𝑠
𝑃𝑇 = 𝑇𝑟𝑎𝑛𝑠𝑝𝑜𝑠𝑒𝑑 𝐼𝑛𝑖𝑡𝑖𝑎𝑙 𝑀𝑎𝑡𝑟𝑖𝑥
From this we can derive the formula
𝐴 = 𝜆1 𝑋1 𝑋1𝑇 + 𝜆2 𝑋2 𝑋2𝑇 + ⋯ + 𝜆𝑛 𝑋𝑛 𝑋𝑛𝑇
Representing the above problem in matrix form
𝐴 = 𝑃𝐷𝑃𝑇
1 2 2 0 0 0 1 2 2
𝐴 = [2 1 −2] [0 3 0 ] [2 1 −2]
2 −2 1 0 0 15 2 −2 1
The orthonormal for eigen vectors can be calculated as
𝑋1 = √(1)2 + (2)2 + (2)2 = √9 = 3
𝑋2 = √(2)2 + (1)2 + (−2)2 = √9 = 3
𝑋3 = √(2)2 + (−2)2 + (1)2 = √9 = 3
 1⁄ 2⁄ 2⁄
3 3 3
2
𝑋1 = ⁄3 1
𝑋2 = ⁄3 𝑋3 = −2 ⁄3
2 −2 1
[ ⁄3] [ ⁄3] [ ⁄3 ]
1⁄ 2⁄
3 3
𝐴 = 0 2⁄3 [1⁄3 2⁄3 2⁄3] + 3 1⁄3 [2⁄3 1⁄3 −2⁄3]
2 −2
[ ⁄3] [ ⁄3]
2⁄
3
+ 15 −2 ⁄3 [2⁄3 −2⁄3 1⁄3]
1
[ ⁄3 ]
8 −6 2
= [−6 7 −4] = 𝐴
2 −4 3
 PSEUDO INVERSE OF A MATRIX

The concept of Pseudo Inverse is also called Moore – Penrose Pseudo inverse
which is used to compute the best fit (least squares) solution to a system of linear
solution that lacks unique solution.
It is denoted by the symbol 𝑨+

To compute the pseudo inverse of a matrix, we must know the concept of

a. Full Row Rank
b. Full Column Rank
c. Rank Deficient System

Full Row Rank: Rank of matrix of order 𝑚 × 𝑛 is equal to the number of rows in
the matrix.

Full Column Rank: Rank of matrix of order 𝑚 × 𝑛 is equal to the number of

columns in the matrix.

Rank Deficient System: Rank of matrix of order 𝑚 × 𝑛 is less than the number
of rows or columns in the matrix.

For these, the governing formulas are

Full Row Rank: If the rows are linearly independent, pseudo inverse is given as
𝐴+ = (𝐴)𝑇 ∙ (𝐴𝐴𝑇 )−1

Full Column Rank: If the columns are linearly independent, pseudo inverse is
given as
𝐴+ = (𝐴𝐴𝑇 )−1 ∙ (𝐴)𝑇

If the system of equations is Rank Deficient

𝐴+ (𝑈Σ𝑉 𝑇 )−1 = (𝑉 𝑇 )−1 Σ −1 𝑈 −1 = 𝑉Σ −1 𝑈 −1
1⁄ 0
𝜎1 0 𝜎1
−1
If Σ = [ 0 𝜎2 ] Σ =[ 0 1⁄ ]
𝜎2
0 0
0 0