Dhaka Univ. J. Sci.

60(2): 209-215, 2012 (July)

A Technique for Solving Special Type Quadratic Programming Problems

M. Babul Hasan
Department of Mathematics, University of Dhaka, Dhaka-1000, Bangladesh
Email: [email protected]
Received on 20. 07. 2011. Accepted for Publication on 22. 02. 2012.
Abstract
Because of its usefulness in production planning, financial and corporate planning, health care and hospital planning, quadratic
programming (QP) problems have attracted considerable research and interest in recent years. In this paper, we first extend the simplex
method for solving QP problems by replacing one basic variable at an iteration of simplex method. We then develop an algorithm and a
computer technique for solving quadratic programming problem involving the product of two indefinite factorized linear functions. For
developing the technique, we use programming language MATHEMATICA. We also illustrate numerical examples to demonstrate our
technique.

Key Words: Quadratic programming, simplex method, algorithm, computer technique

I. Introduction That is, quadratic programming is concerned with the NLPP
Non-linear Programming is a mathematical technique for of maximizing (or minimizing) the quadratic objective
determining the optimal solutions to many business function, subject to a set of linear inequality constraints.
problems. In a non-linear programming problem, either the
Quadratic programming problem
objective function is non-linear, or one or more constraints
have non-linear relationship or both. Let and .Let be a symmetric real
General Non-linear Programming Problem matrix.Then,
Let be a real valued function of variables defined
Maximize
(a) ,
Let be a set of constraints such that subject to the constraints and

where and in a real matrix, is called

(b) a general quadratic programming problem.
Note: represents a quadratic form.The reader may
. recall that a quadratic form is said to be positive-
.
definite (negative definite) if for
and positive-semi-definite (negative-semi-definite)
if for all such that there is one
satisfying .
where are real valued functions of variables It can easily be shown that
.
1. If is positive-semi-definite (negative-semi-
Finally, let
definite) then it is convex (concave) in over all of ,
(c) xj ≥ 0. j = 1, 2, ... ... ... .. m
and
If either or
2. If is positive-definite (negative-definite) then it is
some ; or strictly convex (strictly concave) in over all of .
both are non-linear, then the problem of determining the n-
type which makes a maximum or These results help in determining whether the quadratic
objective function is concave (convex) and the
minimum and satisfies (b) and (c), is called a general non-
linear programming (GNLP) problem. An NLP may be non- implication of the same on the sufficiency of the Karush-
linearly constrained or linearly constrained. If the objective Khun-Tucker conditioin for constrained maxima (minima)
of a linearly constrained NLP is a quadratic function then of .
this problem is called quadratic programming (QP) problem.

*Author for correspondence

210 M. Babul Hasan

Special type Quadratic programming problem Where are defined as in Section

In this chapter, we consider the quasi-concave quadratic
programming problem (QP) subject to linear constraints as Let be the initial basic feasible solution to the (QP).
follows: Then where ,
(QP):Maximize

subject to , Also let , and

where A = is a
In addition we assume that
matrix, b , x, c , . Here we and

assume that Now let be another basic feasible solution where

(i) and are positive for all is the basis in which is
feasible solution. replaced by of but not in .Then the columns of
(ii) The constraints set are given by
is non
empty and bounded. for , and

In general, a quadratic programming problem is solved Then the new value of the basic variables in terms of the
starting with some which satisfies the constraints but does original ones and are as follows:
not necessarily maximize Then is replaced by a series
, and
of other (not necessarily basic feasible solutions) all of
which satisfy the constraints until the required solution is (Say)
obtained.There are many methods for solving (QP). Among
them Wolf’s [4] method, Swarup’s [5] simplex type method Where
and Gupta and Sharm’s [1] 2-basic variable replacement
New optimizing value
method are noteworthy. In this paper, we develop a
computer technique for solving special type (quasi-concave)
of QP problem in which the objective function can be
factorized. Then compare the values of the objective
function at these points in order to choose the optimal
solution. We will introduce a computational technique to
solve the QP problems by using computer algebra
Mathematica [11].
where and
The rest of the paper is organized as follows.
In Section II, we present the simplex type method
of Swarup [5] with numerical example.
In Section III, we present our technique along with
algorithm. We then conclude the paper in Section IV.
Swarup’s Simplex Type Method for Solving QP

Adding slack variables to the

constraints of (QP), we have the standard (QP) as
follows:
Similary,
(QPI): Maximize
Hence
subject to ,
A Technique for Solving Special Type Quadratic Programming Problems 211

at a simplex table, the solution becomes

optimal and the process terminates.
Criterion1: (Choice of the entering variable)
Optimality condition To identify the entering variable, choose the uth column of
A for which is the greatest positive ; j = 1, 2,…, n
The value of the objective function will improve If
Criterion 2: (Choice of the outgoing variable)

Choose for which

Numerical Example
Maximize
But for non degenerate case

subject to , ,
,
where
Hence we must have, Solution
Introduction the slack variables to the constraints we get the
Thus the solution can be improved until for all initial table as follows:
Whenever all for all

Table. 1. (Initial table)

2 4 1 0 0 0
1 1 2 0 0 0 b

0 0 1 0 1 0 0 4
0 0 2 3 0 0 1 0 3
0 0 0 1 4 0 0 1 3
1

2 4 1 0 0 0
1 1 3 0 0 0

3/2 4/3 ¾ - - -
8 43/3 11/2 - - -
212 M. Babul Hasan

Table.2
2 4 1 0 0 0
1 1 2 0 0 0 b

4 1 1/3 1 0 1/3 0 0 4/3

0 0 5/3 0 0 -1/3 1 0 5/3
0 0 -1/3 0 -1/3 0 1 5/3
4

2/3 0 1 -4/3 0 0
2/3 0 2 -1/3 0 0

1 - 5/12 4 - -
62/9 - 101/6 -43/9 - -

Table.3
2 4 1 0 0 0
1 1 2 0 0 0 b

4 1 1/3 1 0 1/3 0 0 4/3

0 0 0 0 -1/3 1 0 5/3
0 0 -1/12 0 1 -1/12 0 1/4 5/1
5/3

3/4 0 0 -5/4 0 -1/4

5/6 0 2 -1/6 0 -1/2

1 - - 4 - -
75/8 - - 11/2 - 101/24
-

Table.4
2 4 1 0 0 0
1 1 2 0 0 0 b

4 1 0 1 0 2/5 -1/5 0 1
2 1 1 0 0 -1/5 3/5 0 1
1 2 0 0 1 -1/10 1/20 1/4 1/2

0 0 0 -11/10 -9/20 -1/4

0 0 0 0 -1/2 -1/2

- - - 5/2 5/3 2
- - - -11/2 -45/8 -19/4
A Technique for Solving Special Type Quadratic Programming Problems 213

Since all is non positive in Table-4, it gives the global Step 3: Test whether the linear system of equations has
unique solution or not.
optimal solution.
Step 4: If the system of linear equations has got any unique
Hence the optimal solution is
solution, find it.
with global
Step 5: Dropping the solutions with negative elements,
maximum, . determine all basic feasible solutions.
III. Algorithm for Solving Special Type QP Step 6: Calculate the values of the objective function for
In this section, we present the computational technique the basic feasible solution found in step 5.

for solving the QP problems. In 2000, Hasan et. al. [10] Step7: For the maximization of QP the maximum value of
developed a computer oriented solution method for solving is the optimal value of the objective function, and the
the LP problems. In this study, we will extend that method basic feasible solution which yields the optimal value is the
for solving the QP problems. We will see that, this method optimal Solution.
will help us to solve QP problems easily and to avoid the
clumsy and time consuming methods of Swarup [5] and Computer code for solving special type QP
other methods. In this section, we present our method for solving QP
Step 1: Express the QP to its standard form. problems using computer algebra Mathematica.

Step 2: Find all sub-matrices of the coefficient

matrix by setting variables equal to zero.
214 M. Babul Hasan

INPUT:

OUTPUT:

Graphical representation gives the actual concept of the

problems. As, this is an optimization problem, we present
graphical representation of this problem in the next section
in order to achieve a more visual concept of the solutions
of the given problem.

Graphical representation of numerical example of

Section 2:
A Technique for Solving Special Type Quadratic Programming Problems 215

coefficient matrix A, right hand side constant b, cost

coefficient vectors c and d and the constants  and  in
the same program and easily obtained the optimal solution.
Our method is also able to solve degenerate and cycling
problems easily whereas the simplex method is proved to
be inconvenient for solving such problems.
IV. Conclusion
In this paper, we developed a computer technique for
solving special type (Quasi-Concave) of quadratic
programming in which the objective function can be
factorized. We illustrated the solution procedure developed
by us with a numerical example. We observed that the
result obtained by our procedure is completely identical
with that of the other methods which are laborious and time
consuming and clumsy. We therefore, hope that our
program can be used as an effective tool for solving special
type QP problems in which the objective function can be
factorized and hence our time and labor can be saved.
…………..
1. Swarup, K., P. k. Gupta and Man Mohan, 2003,”Tracts in
Operations Research”, Eleventh Thoroughly Revised
Edition,ISBN:81-8054-059-6
2. Frederick S. Hillier and Gerald J. Lieberman, McGraw-Hill
International Edition 1995, ” Introduction to Operations
Research”
3. Wayne L. Winston, (1994), “Application and algorithms,”
Dexbury press, Belmont, California, U.S.A.
4. Wolfe, P., (1959), “The simplex method for quadratic
programming,” Econometrica, 27(3), 382-398.
5. Swarup, K. 1966, “Quadratic Programming” CCERO
(Belgium), 8(2), 132-136.
6. Swarup, K., 1964, “Linear fractional functional
programming,” Operations Research, 13 (6), 1029-1036.
7. Dantzig, G.B., 1962, “Linear programming and extension,”
Princeton University press, Princeton,N.J.
8. Hasan, M.B. and M.A.Islam, 2001, “Generalization of
simplex method for solving fractional programming
problems,” The Dhaka University Journal of Science, 53(1),
111-118.
9. Hasan, M.B. (2008), “Solution of Linear Fractional
Programming Problems through Computer Algebra”, The
Dhaka University Journal of Science, 57(1),
10. Hasan, M. B., A.F.M.K Khan and M. A. Islam, 2001, “Basic
solutions of linear system of equations through computer
We observed that with Swarup’s [5] method we algebrs,” Ganit, Journal of Bangladesh Mathematical Society,
21 (1), 39-45.
experienced complex calculations for the simplex tables.
For solving problems with “  type” or “= type” 11. Wolfram, S. 2000, “Mathematica,” Addision-Wesly
constraints, one needs to solve that problem by Two-phase Publishing Company, Melno Park, California, New York.
or Big-M simplex method which is clumsy and time
consuming. But by our method, we just had to compute the