You are on page 1of 22

Karmarkar’s Interior Point Method

Kwame Nkrumah University of Science and Technology,


College of Science
Deparment of Mathematics
Professor J. Ackorah Prah
July 24, 2023
Group 13

Members
Wiafe Maxwell 9338519
Appiah Ernest 9326319
Appiah Eugene 9326419
Tayie Clement 9337519
Tiigah Patrick Wonbod 9337719
Agyeman Agyapomaa Augustina 9323919
Ahenkorah Maxwell 9324419
Nuamah Kwabena Tutu 9334819
Osei James Bruce 9335819
Giba Jeffery Amidima 9331419

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 2 / 22
Table Of Contents

1 Introduction
2 Methodology- Algorithm and worked example
3 Advantage and Disadvantage
4 Application
5 Conclusion
6 Reference

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 3 / 22
Introduction

The simplex method of linear programming finds the optimum


solution by starting at the origin and moving along the adjacent
corner points of the feasible solution space.
Since this is exponential time algorithm, the numbers of iterations
become prohibitive for some huge problems .

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 4 / 22
Introduction Cont.

The interior point Linear Programming algorithm is a polynomial time


algorithm that finds an optimum solution by starting at a trial solution
and shooting through the interior of the feasible solution space
It is ideal for solving very large LP problems.
It can also be used to solve nonlinear convex optimization problems
It is also called the barrier method.

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 5 / 22
The Basic Idea Of Karmarkar’s Algorithm

The performance of an iterative algorithm depends upon two key factors:


1 How many steps (iterations) does it take?
2 How much computation does it involve in each iteration?
The simplex method starts with an extreme point and keeps moving
to a better neighboring extreme point at each iteration until an
optimal solution or infeasibility is reached.
It means that the scheme is designed to reduce the amount of work
done during each iteration by only considering the directions that lead
to the nearest extreme points.

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 6 / 22
Cont.

But, as the Klee-Minty example showed, the simplex method may


have to travel a long path on the boundary of the feasible domain and
visit almost every extreme point before it stops. This boundary
approach suffers from heavy computation in large-scale applications,
since the feasible domain may contain a huge number of extreme
points.
Therefore one alternative idea is to travel across the interior of the
feasible domain along a"shorter path" in order to reduce the total
number of iterations. However, this interior point approach usually
requires the consideration of all feasible directions for a better
movement at each iteration.

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 7 / 22
Cont.

In general, it can be difficult to identify the best direction of movement


among all feasible directions at a particular interior point of the feasible
domain. However, Karmarkar noticed two fundamental insights that can
help, assuming the feasible domain is a polytope.
1 If the current interior solution is near the center of the polytope, then
it makes sense to move in the direction of steepest descent of the
objective function to achieve a minimum value.
2 Without changing the problem in any essential way. an appropriate
transformation can be applied to the solution space such that the
current interior solution is placed near the center in the transformed
solution space.

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 8 / 22
Diagram

This insight can be observed as follows;

[Fig 1]

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 9 / 22
Karmarker’s Standard Form

Folowing the basic strategy of the projective scaling. Karmarkar’s


algorithm has a preferred standard form for linear programming:
Minimize

cT x (1)

Subject to

Ax = 0 (2)

e T x = 1.x ≥ 0 (3)
where A is an m x n dimensional matrix of full row rank. e T = (1, 1...1) is
an n-vector of all ones, c is the coefficient of the objective function,
x ∈ R n.

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 10 / 22
The Simplex Structure

The key karmarkar’s standard form is the Simplex Structure, which of


results in bounded feasible domain. So equation (3) defines the simplex
structure
n
X
△ = x ∈ R n| xi = 1, x1 ≥ 0
i=1

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 11 / 22
Karmarker’s assumption
Karmarkar made two assumptions for his algorithm, these are;
e
1 Ae = 0 so x 0 = n = ( n1 , ..., n1 )T is an initial interior solution.
2 The optimal objective value of problem is 0.

Key Parameters
x k = interior feasible solution
Xk = n dimentional diagonal matrix with ith elements of vector x k
Bk = contraints matrix of LP problem in Karmarkar’s standard form
dk = feasible direction
y k+1 = new interior feasible solution in the transformed space.
x k+1 = new interior feasible solution

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 12 / 22
Procedure For The Implementation Of Karmarkar’s
Algorithm

e
Initialization; for x k set k = 0, x 0 = n and L to be a larger positive
integer.
Optimality check: If c T x k ≤ 2−L (c T ne ) then stop with an optimal
solution x ∗ = x k otherwise go to step 3.
Find a better solution; Let

Xk = diag(X k )

AXk
Bk = [ ]
eT

d k = −[I − BkT (Bk BkT )−1 Bk ]Xk c

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 13 / 22
Cont...

e α dk
y k+1 = + ( k )
n n ∥d ∥

Xk y k+1
X k+1 =
e T Xk y k+1
set k+1: go to step 2

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 14 / 22
Worked Example

Minimize

−x1 + 1
Subject to the constraints

x2 − x3 = 0

x1 + x2 + x3 = 1

x1 , x2 , x3 ≥ 0

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 15 / 22
Solution the problem

first we see that the linear programming problem is in Karmarkar’s


standard form which satisfies the Karmarkar’s assumptions, hence we start
with x 0 = ( 13 , 31 , 31 )
h andi note that
A= 0 1 −1 and c T = (−1, 0, 0).
Now check step 2 and choose L = 20 and easily see that the objective
value x 0 is too high. Therefore we must find a better solution.
Step 3. We define 
1
3 0 0
x0 =  0 13 0 
 
0 0 31

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 16 / 22
Cont...(Solution the problem)
" #
AX0
then AX0 = [0, 31 , −1
3 ] and
eT
" #
−1
0 31 3
B0 =
1 1 1
Moveover, the moving direction is given by
2 −1 −1 T
d 0 = −[I − B0T (B0 B0T )−1 B0 ]X0 c = ( , , )
9 9 9
√ √
With norm ∥d∥ = 96 for the purpose of illustration we choose α =1/ 6
to obtain a new solution in the transformed space

e α d0
y1 = + ( 0 )
n n ∥d ∥

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 17 / 22
Cont...(Solution the problem)

1 1 2 −1 −1 T 9
y1 =
+ √ ( , , ) (√ )
3 3 6 9 9 9 6
4 5 5
= ( , , )T
9 18 18
Hence, the new interior feasible solution is given by

X0 y 1 4 5 5
x1 = T 1
= ( , , )T
e X0 y 9 18 18
Continuing this iterative process, Karmarkar’s algorithm will stop at the
optimal solution x ∗ = (1, 0, 0)T . It is worth mentioning that if we take
α = √66 > 1, then y 1 = (1, 0, 0)T and x 1 = x ∗ .
Hence, direction d 0 really points to the optimal solution

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 18 / 22
ADVANTAGES AND DISADVANTAGES

ADVANTAGES
It is efficient in solving large-scale LP problems with reasonable
numbers of iteration.
This method can handle degenerate LP problems, where multiple
optimal solutions exist at the corner point of the feasible region.

DISADVANTAGES
This requires specialized software or programming skills making it less
accessible to users who are not familiar with numerical optimization
techniques.
If an inappropriate initial point is selected the algorithm will take more
iterations to converge or may converge to a sub optimal solution.

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 19 / 22
Application

Operations Research Management


Finance and Investment
Supply Chain Management
Network Flow Management
Energy Production and Distribution

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 20 / 22
Conclusion

The Karmarkar interior-point method is an efficient optimization algorithm


used to find solutions to linear programming problems. It converges rapidly
to the optimal solution by exploring the interior of the feasible region,
making it well-suited for large-scale problems. By iteratively updating the
solution, Karmarkar’s method efficiently handles both equality and
inequality constraints, delivering accurate results with a reduced number of
iterations compared to other methods.

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 21 / 22
Reference

Anstreicher. K. M.. "A combined phase I- phase lI projective algorithm for


linear program- ming." Mathematical Programming 43. 209-223 (1989).
Anstreicher. K. M.. n O" the performance of Karmarkar’s algorithm over
asequence of iterations."SIAM Journal on Optimization 1. 22-29(1991).
Bayer. D. and Lagarias. .J C. "Karmarkar’s linear programming algorithm
and Newton’s

(Bsc. Mathematics(Group 13)) Karmarkar’s Interior Point Method July 24, 2023 22 / 22

You might also like