Separable Programming Presentation
Separable Programming Presentation
MA7391 Seminar
Separable Programming
By
VIYASAR MOULY
(M190659MA)
Department of Mathematics
National Institute of Technology Calicut
OVERVIEW
• INTRODUCTION
• SEPARABLE PROGRAMMING
• REFERNCE
Introduction
Separable function
Definition 1.1
A function f (x1 , x2 , ....xn ) is said to be separable if it can be
expressed as the sum of n single valued functions f1 (x1 ), f2 (x2 ), ....f1 (xn ), i.e.
Separable programming
Definition 1.2
If the objective function of NLPP problem can be expressed as a
linear combination of several different one-variable functions,of which same
or all are non-linear,Then such an NLPP problem is called a Separable
programming.
Consider,
subject to
xi ≥ 0 i,j=0,1,2,...n
Since the objective functions as well as all the constraints are separable.
EXAMPLES
is separable fumction.
• The function
• Some non linear functions are not directly separable but can be made
separable by appropriate substitutions.
Case I
Max Z=y
subject to
ln y=lnx1 +lnx2
Which is separable.
Case II
w1 = x1 + ∂1
w2 = x2 + ∂2
w1 , w2 > 0
Now,
x1 x2 = (w1 − ∂1 )(w2 − ∂2 )
x1 x2 =w1 w2 − w2 ∂1 − w1 ∂1 + ∂1 ∂2
continued..
Let y=w1 w2 ,
Maximize Z=y-∂1 w2 − ∂2 w1 + ∂1 ∂2
subject to
ln y=ln w1 +ln w2
Which is separable.
Definition
It is the special case of Separable programming in which the
separable objective function(in minimizing form) all separable constraints
(≤ type) are convex.
n
X
= fj (xj )
j=1
subject to
gi (x) ≤ b
n
X
= gij (xj ) ≤ bi
j=1
i=1,2,...m. xj > 0,
j=1,2,...n.
Here,λk + λk+1 = 1
continued
•Let we take x is sub interval[xk , xk+1 ] and [xk+1 , xk+2 ]
•Any x can be express the linear combination of the ,
k
X
λk = 1
r=0
0 0
• Atleast one or no more than two λks are positive.Two λks positive ,they must
be concecutive.
Viyasar Mouly(M190659MA) (NIT Calicut) Separable programming 16:20, December 3, 2020 16 / 33
SEPARABLE PROGRAMMING
conitnued...
(x − xk )
f(x)=f (xk ) + f (xk+1 ) − f (xk )
(xk+1 − xk )
Adjacent criteria
where,
k
X
λk = 1
r=0
0
Atmost two λks are greater than zero.
they must be concecutive.
Objective function
Xn
Max/min z= fi (xi )
i=1
subject to
Xn
gij (xij ) ≤ bj i,j=12,..n.
i=1
xi ≥ 0 i=1,2,...,n.
li ≤ xi ≤ xi
Constraints
The constraints can be expressed th linear combination of
ri
X
gij (xij ) = λik gijk (xik )
k=0
subject to
ri
n X
X
(λik )(gijk ) ≤ bj j=1,2,3....n.
i=1 k=0
Xri
λik = 1 0 ≤ λik ≤ 1
i=1
k=0,1,2,....ri
i=1,2,3....n.
fj andgij are also separable functions.
Adjacents critria λ-form equivalent to LPP.
Viyasar Mouly(M190659MA) (NIT Calicut) Separable programming 16:20, December 3, 2020 20 / 33
SEPARABLE PROGRAMMING
Algorithm
Step 1
If the objective function is of minimization form, convert it into maximization.
Step 2
Test whether the functions fj (xj )andgij (xj ) satisfy the concavity (convexity)
conditions required for the maximization(minimization) of non-linear
programming problem. If the condition are not satisfied, the method is not
applicable, otherwise go to next step.
Algorithm
Step 3
Divide the interval 0 ≤ xj ≤ tj (j = 1, 2, ..., n) into a number of mesh points
ajk (k = 1, 2, ..., Kj ) such that aj1 = 0,aj1 < aj2 < ... < ajk = tj
Step 4
For each point xjk , compute piecewise linear approximation for each fj (xj ) and
gij (xj ) (j = 1,2,..., n; i = 1,2,...,m) .
Algorithm
Step 5
Using the computations ofstep - 4, write down the piecewise linear
approximation ofthe given NLPP.
Step 6
Solve the resulting LPP by two phase Simplex method. In this method, treat
λj0 (j = 1,2,...,ri) as the artificial variables and since the costs associated with
them are not there, these are assumed to be zeroes, the phase -1 of the two -
phase method is automatically completed. Thus the initial simplex table of
phase -1 becomes optimum and hence represents the starting simplex table for
phase - II
Algorithm
Step 7
Obtain the optimum solution x0 ofthe given NLPP by using the relations
X m
0
x = λjk (xjk ) i, j = 12, ..n.
k=0
and find optimum value ofthe objective function as follows
Xn
0
f (x) = f j (xj0 ) i,j=12,..n.
j=1
PROBLEM
Solve minimize x12 − 4x1 − x2
Subjext to 2x12 + 8x2 ≤ 30
0≤ x1 ≤ 2
0≤ x2 ≤ 2
here
f (x1 ) = x12 − 4x1 f (x2 ) = −x2
g(x1 ) = 2x12 g(x2 ) = 8x2
problem
Minimize
n X
X n
λij fij =−3λ11 − 4λ12 − λ21 λ22
i=1 j=1
Subject to
2λ11 + 8λ12 + 8λ21 + 32λ22 ≤ 30
λ10 + λ11 + λ12 = 1
λ20 + λ21 + λ22 = 1
λ11 , λ12 , λ21 , λ22 , λ10 , λ20 ≥ 0
Phase I
• In this method
λ11 , λ12 , λ21 , λ22 , λ10 , λ20
as the artificial variables and since the costs associated with them are not
there, these are assumed to be zeroes, the phase -1 of the two - phase method
is automatically completed.
Phase II
Table 1
Pivot elemet 1,The non basic variableλ12 enter the basic and the basic
variable λ10 leaves the basis.
Viyasar Mouly(M190659MA) (NIT Calicut) Separable programming 16:20, December 3, 2020 28 / 33
SEPARABLE PROGRAMMING
Table 2
Pivot elemet 1,The non basic variableλ21 enter the basic and the basic
variable λ20 leaves the basis.
Table 3
Pivot elemet 32,The non basic variableλ22 enter the basic and the basic
variable x3 leaves the basis.
Table 4
Reference