Module1 1

Lecture-4

Basic Feasible Solution:

Consider a system of linear equation
Ax = b, A = m × n = [aij ], b : m × 1; x : n × 1. The matrix [A, b]
is called augumented matrix. The necessary and sufficient condition for a system to be
consistent is that ρ(A) = ρ([A, b]).
Further, if ρ(A) = ρ([A, b]) = n(no. of unknown) then the system has a unique solution,
while if ρ(A) = ρ([A, b]) < n then the system has infinite solutions.

Example 1.15
x1 + x2 + x3 = 1
x1 + 2x2 = 3
3x1 + x2 − x3 = 5
 
1 1 1 1
[A, b] =  1 2 0 3 
 

3 1 −1 5
R2 → R2 − R1 , R3 → R3 − 3R1
 
1 1 1 1
=  0 1 −1 2 
 

0 −2 −4 2
1
R3 → − R3 1
 2 
1 1 1 1
=  0 1 −1 2 
 

0 1 2 1

Now ρ(A) = 3 = ρ([A, b]) = n. So the system has a unique solution i.e.
x1 = 1, x2 = 1, x3 = −1

Consider a system Ax = b of m equations in n unknown(n > m). Let ρ(A) = ρ([A, b]) =
m, i.e., none of the equations is redundant.

Definition1.13 Basic Solution: A solution obtained by setting exactly n − m variables

to zero provided the determinant formed by the columns associated to the remaining m
variables is non zero is called Basic Solution.

Copyright
c Reserved IIT Delhi
 2
The remaining m variables are" termed
# as basic variables
xB
Ax = b ⇒ [B, N ] = b, |B| =
6 0 (B-basis matrix)
xN
xN = 0 ⇒ BxB = b ⇒ xB = B −1 b (xN -non basic variables).

Thus a solution in which the vectors associated to m variables are L.I. and remaining
n − m variables are zero is called a basic solution. Note that for a solution to be basic,
atleast n − m variables must be zero.

Example 1.16 Find(Graphically) basic feasible solution of

2x1 + 3x2 ≤ 21
3x1 − x2 ≤ 15
x1 + x2 ≥ 5
x2 ≤ 5
Feasible se
x 1 , x2 ≥ 0
2
x
x2
X
|

x2*isX not integer

|

(2) x x !
|

1 x2* +1
└ x┘
|

B(3,5)
(4) x
(0,5)
|

x2*
└ ┘
|

(6,3) Can be removed

x
|

(3)
|

(1) y
|

| | | | | | | | | | |
x1
O
(5,0)

(0, 5) → (0, 5, 6, 20, 0, 0) → degenerate BFS

(5, 0) → (5, 0, 11, 0, 0, 5) → degenerate BFS
(6, 3) → (6, 3, 0, 0, 4, 2) → nondegenerate BFS
(3, 5) → (3, 5, 0, 4, 3, 0) → nondegenerate BFS

Example 1.17 Find all basic solutions of the system

2x1 + 6x2 + 2x3 + x4 = 3
6x1 + 4x2 + 4x3 + 6x4 = 2
Solution: α1 x1 + α2 x2 + α3 x3 + α4 x4 = b
" # " # " # " # " #
2 6 2 1 3
α1 = , α2 = , α3 = , α4 = , b=
6 4 4 6 2
 
0
B1 = (α1 , α2 ), |B1 | = −28 6= 0 ⇒ xB1 = 1 
2

Copyright
c Reserved IIT Delhi
   3
−2
B2 = (α1 , α3 ), |B2 | = −4 6= 0 ⇒ xB2 =  7 
 2 
8
 3 
B3 = (α1 , α4 ), |B3 | = 6 6= 0 ⇒ xB3 =  −7 
 3
1
B4 = (α2 , α3 ), |B4 | = 16 6= 0 ⇒ xB4 = 2 

0
 
1
B5 = (α2 , α4 ), |B5 | = 32 6= 0 ⇒ xB5 =  2 
0
" #
2
B6 = (α3 , α4 ), |B6 | = 8 6= 0 ⇒ xB6 =
−1

Basic Solution:-
 
1
0 0 0 , - Degenerate Basic Feasible Solution
 2    h
−7 8 −7
i
−2 0 0 , 0 0 , 0 0 2 −1 −Nondegenerate Basic Solutin
3 3 3

Example 1.18 Basic solutions of the system

x1 + 2x2 + x3 = 4
are
2x1 + x2 + 5x3 = 5

x1 = 2, x2 = 1, x3 = 0
x1 = 5, x2 = 0, x3 = −1
5 2
x1 = 0, x2 = , x3 =
3 3
All basic solutions are degenerate.

Number of Basic Solution:

If Ax = b, A : m × n, ρ(A) = m then maximum number of basic solutions is n Cm .

In L.P.P. a feasible solution which is also basic is called a basic feasible solution. Recall
that feasible solution satisfies the set of constraints and the non-negativity restriction.

A feasible solution in which n − m variables are zero and the vectors associated to the
remaining m variables, called basic variables, are Linearly Independent, is called a B.F.S.

Obviously a feasible solution which contains more than m positive variables is not a basic
feasible solution.

Definition1.14 Degenerate Basic Solution: If any of the basic variables vanishes,

the solution is called degenerate basic solution. On the other hand if none of the basic

Copyright
c Reserved IIT Delhi
 4
variables vanishes, the solution is called non-degenerate basic solution. Thus, a non-
degenerate basic solution contains exactly m non-zero and n − m zero variables.

Theorem 1.7 Every extreme point of a convex set of all feasible solution of the L.P.P.
Max z = cx
subject to Ax = b
x ≥ 0
is a basic feasible solution and vice-versa.

Example 1.19 Which of the following vectors is a basic feasible solution of system
x1 + 2x2 + x3 + 3x4 + x5 = 9
2x1 + x2 + 3x4 + x6 = 9
−x1 + x2 + 3x4 + x7 = 0
xi ≥ 0
x1 = (2, 2, 0, 1, 0, 0, 0), x3 = (3, 3, 0, 0, 0, 0, 0)
x2 = (0, 0, 9, 0, 0, 9, −9), x4 = (0, 0, 0, 0, 9, 9, 0)
x5 = (1, 0, 0, 0, 8, 7, 1), x6 = (0, 0, 0, 3, 0, 0, 0)
Solution.(i) Columns  associated
 with  non-zero
 variable
 inx1 are
1 2 3
a1 =  2  , a 2 =  1  , a 4 =  3 
     

−1 1 0
There exist scalars 1,1,-1 such that
1.a1 + 1.a2 − 1.a4 = 0 ⇒ Linearly Dependent
⇒ x1 is not basic feasible solution.

(ii) Vectors associated with non-zero variables are a3 = (1, 0, 1)T , a6 = (0, 1, 0)T ,
a7 = (0, 0, 1)T .
1 0 0
|B| = 0 1 0 = 1 6= 0
1 1 1
⇒ vectors a3 , a6 , a7 is Linearly Independent.
⇒ x2 is a basic solution but not an basic feasible solution as x7 < 0.
   
1 2
(iii) Vectors associated with non-zero variables are a1 =  2  , a2 =  1 
   

−1 1
 
1
Now these vectors together with vector a5 =  0  are Linearly Independent as
 

0
1 2 1
|B| = 2 1 0 = 3 6= 0
−1 1 0

Copyright
c Reserved IIT Delhi
 5
This solution is a basic feasible solution taking x1 , x2 , x3 as basic variables. Degenerate
B.F.S. as x5 = 0.

(iv) Columns a5 , a6 together with a7 are Linearly Independent. So solution x4 is a

degenerate B.F.S. as basic variable x7 = 0.

(v) The solution x5 contains more than 3 non-zero variables so it is not a B.F.S.
 
3
(vi) There is only one nonzero variable with column a4 =  3 
 

0
This vector along with a6 and a7 is L.I. as
3 0 0
|B| = 3 1 0 = 3 6= 0
0 0 1
This Solution is a degenerate B.F.S..

Fundamental Theorem of Linear Programming

Theorem 1.8 If there is a feasible solution to the system Ax = b, x ≥ 0, where A is
m × n matrix, m < n, ρ(A) = m, then there is also a B.F.S.
Proof: Let x = (x1 , x2 , . . . , xn ) be a feasible solution of given system of equations Ax = b,
x ≥ 0.
Suppose out of these n components of x, k are non-zero and rest (n−k) are zero. Without
loss of generality, we assume that first k components of x are non zero. Thus, we have
x = (x1 , x2 , . . . , xk , 0, . . . , 0).
Since x is a feasible solution, we have
k
X
Ax = b ⇒ aj x j = b (1)
j=i
and x ≥ 0 ⇒ xj > 0, ∀ j = 1, 2, . . . , k
Now if {a1 , a2 , . . . , ak } are L.I. then x is a B.F.S.
Suppose this is not the case, i.e., the vectors are L.D. Then there exist scalars λ1 , λ2 , . . . , λk ,
not all zero, such that
X k
λj aj = 0.
j=i
Suppose λr 6= 0. Then we get
k
X λj
ar = − aj = 0.
j=ij6=r
λr
Substituting value of ar in (1), we get
k  
X λj
xj − xr aj = b.
j=ij6=r
λr

Copyright
c Reserved IIT Delhi
 6
k
X
⇒ x̂j )aj = b.
j=1
λj
Now, x̂j = xj − xr , j 6= r.
λr
x̂ = (x̂1 , x̂2 , . . . , x̂k , 0, . . . , 0) is a feasible solution of the system Ax = b, x ≥ 0 if x̂ ≥ 0.
That means
λj
xj − x≥ 0r, ∀j = 1, . . . , k, j 6= r.
λr
λj
Now, if λj ≤ 0 then since xj > 0, xr > 0, λr > 0 we get x̂j = xj − xr ≥ 0. The only
λr
cases are for those λj ≥ 0.
λj xj xr
Then, xj − xr ≥ 0 ⇒ ≥ .
λr λj λr
xr xj
Thus, = min{ : λj > 0} (2)
λr j λj

If we choose r according to rule (2) then x̂j ≥ 0 ∀ j = 1, . . . , k. So, x̂ is a feasible

solution of the given system with atmost (k − 1) non-zero components. If the columns of
A associated with these (k − 1) non-zero components is linearly independent then x̂ is a
B.F.S.
Else we can continue with x̂ in the same manner as done for x to get another feasible
solution x̂ˆ with atmost (k − 2) non-zero components.
In atmost (k − 1) steps we get a solution x̂ with atmost 1 non-zero component and this
is a B.F.S.

Example 1.20 If x1 = 2, x2 = 3, x3 = 1 be a feasible solution of the LPP

Max z = x1 + 2x2 + 4x3
subject to 2x1 + x2 + 4x3 = 11
3x1 + x2 + 5x3 = 14
x 1 , x2 , x3 ≥ 0
then find the basic feasible solution.
Solution.the constraints can be written as
a1 x2"+ a2#x2 + a3 x"3 = # b, xi ≥ 0" # " # (1)
2 1 4 11
a1 = , a2 = , a3 = , b=
3 1 5 14
Since the vectors {a1 , a2 , a3 } are linearly dependent, we get
⇒ x1 is not basic feasible solution. 1a1 + 2a2 − a3 = 0
i.e.λ1 = 1, λ2 = 2, λ3 = −1.
Select r such that  
xr xi
= min , λi > 0
λr  λi 
x1 x2
= min ,
λ1 λ2

Copyright
c Reserved IIT Delhi
   7
3
= min 2,
2
x2
=
λ2
⇒ r = 2
a1 a3
⇒ a2 = − +
2 2
Substitute in (1), we get  
−a1 a3
a1 x 1 + + x 2 + a3 x 3 = b
2 2
 x2  x2 
⇒ x1 − a1 + x 3 + a3 = b
 2   2
3 3
⇒ 2− a1 + 1 + a3 = b
2 2
1 5
⇒ a1 + a3 = b
2 2
1 5
⇒ x1 = , x2 = 0, x3 = is a feasible solution which is also a B.F.S. as {a1 , a3 } is
2 2
linearly independent because
2 4
= 10 − 12 = −2 6= 0.
3 5

Example 1.21 Consider the system of constraints

x1 ≥ 6
x2 + x3 ≥ 2
x 1 , x 2 , x3 ≥ 0
The point x̄ = (7, 2, 0)T is a feasible solution of the system, and the set of column vectors
corresponding to positive xj in the system is {(1, 0)T , (0, 1)T } which is linearly indepen-
dent. However x̄ is not a B.F.S. of this system. If we introduce the slack variables.
x1 − s1 = 6,
x2 + x3 − s2 = 2,
x1 , x2 , x3 , s1 , s2 ≥ 0
and (x̄, s̄) = (7, 2, 0, 1, 0)T

Copyright
c Reserved IIT Delhi