Scribd
Upload
50 views3 pages

Unrestricted and Other Variable Types

The document discusses handling unrestricted and other variable types in linear programs. It describes converting a problem with nonpositive and unrestricted variables into an equivalent problem with all nonnegative variables. For the unrestricted variable x3, two new nonnegative variables x3+ and x3- are introduced such that x3 = x3+ - x3-. This allows solving the problem using the two-phase method. After solving the converted problem, the original optimal solution can be easily reconstructed.

Uploaded by

Rovick Tarife
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
50 views3 pages

Unrestricted and Other Variable Types

The document discusses handling unrestricted and other variable types in linear programs. It describes converting a problem with nonpositive and unrestricted variables into an equivalent problem with all nonnegative variables. For the unrestricted variable x3, two new nonnegative variables x3+ and x3- are introduced such that x3 = x3+ - x3-. This allows solving the problem using the two-phase method. After solving the converted problem, the original optimal solution can be easily reconstructed.

Uploaded by

Rovick Tarife
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Unrestricted and Other Variable Types

Consider the linear program:

Minimize 2x1 +3x2 −x3


Subject to:
x1 −x2 +2x3 ≤ 10 (1)
−3x1 +2x2 −4x3 = 6 (2)
x1 +9x2 ≥ 7 (3)
x1 ≤ 0, x2 ≥ 0, x3 is unrestricted.

In this problem, the variable x1 is nonpositive and the variable x3 is unrestricted in sign.
Before the launch of the two-phase procedure, we need to first convert the problem into one
where all variables are nonnegative.

Handling nonpositive variables is quite easy. The idea is to define a new variable that equals
the negative of the original variable. In this case, we can denote the new variable by x01
(say) and let x01 ≡ −x1 . Since the original variable is nonpositive, it follows that the new
variable x01 is nonnegative. To implement this change of variable, we simply replace every
instance of x1 in the given problem by −x01 .

More generally, a variable may sometimes have a given constant upper bound. For example,
suppose the requirement x1 ≤ 0 is revised to x1 ≤ 3. Observe that the latter can be rewritten
as x1 − 3 ≤ 0, which suggests that if we define x̄1 ≡ x1 − 3, then x̄1 is a nonpositive variable.
Hence, we can replace every instance of x1 in the given problem by x̄1 + 3 to create an
equivalent problem with x̄1 ≤ 0. Notice that doing so will create an extra constant in the
objective function. The constant will not have any impact on the Simplex algorithm, but
it is recommended that it be set aside temporarily until the end of the solution procedure.

Of course, a variable with a constant lower bound, say x1 ≥ 5, can be handled similarly:
Replace every instance of x1 by x̄1 + 5, where x̄1 ≥ 0.

We now turn our attention to the handling of unrestricted variables, x3 in this case. The
idea is to rewrite such a variable as the difference between two nonnegative variables. That
is, we can introduce two nonnegative variables, say x+ − + −
3 and x3 , and let x3 ≡ x3 − x3 .

To see how this works, consider a numerical example. Suppose the value of x3 is to equal 5;
then, the assignments x+ −
3 = 8 and x3 = 3 will yield a difference of 5. Such assignments are
clearly not unique. Other examples are: x+ − + −
3 = 18 and x3 = 13; and x3 = 11 and x3 = 6.
In fact, such assignments are innumerable.

1
Of particular interest is the assignments x+ −
3 = 5 and x3 = 0. Notice that in this case, we
have x+ −
3 = max[x3 , 0] and x3 = − min[x3 , 0]. When these two relations hold, we say that
x+ −
3 is the positive part of x3 and x3 is the negative part of x3 . Thus, if the variable x3
equals 5, then its positive part is 5 and its negative part is 0.

As another numerical example, suppose x3 is to equal −5. Then, the positive part of x3 ,
denoted by x+ −
3 , equals 0; and the negative part of x3 , denoted by x3 , equals 5. Indeed,
x3 = 0 − 5 = −5; and both the positive part and the negative part are nonnegative.

We now execute the proposed conversions for x1 and x3 . With x1 replaced by −x01 and x3
replaced by x+ −
3 − x3 , the given problem is equivalent to the linear program below.

Minimize −2x01 +3x2 −x+


3 +x−
3
Subject to:
−x01 −x2 +2x+
3 −2x−
3 ≤ 10 (1)
+
3x1 +2x2 −4x3 +4x−
0
3 = 6 (2)
0
−x1 +9x2 ≥ 7 (3)
x01 ≥ 0, x2 ≥ 0, x+ −
3 ≥ 0, x3 ≥ 0 .

Thus, the new problem has 4 variables, all of which are nonnegative; and it is now ready
for the application of the two-phase procedure.

After solving the new problem, it is very easy to convert its optimal solution into one for
the original problem. As an example, consider an arbitrarily chosen feasible solution to
the new problem, say (x01 , x2 , x+ − 0 +
3 , x3 ) = (1, 1, 1, 1). Since x1 = −x1 and x3 = x3 −
x−
3 , the corresponding feasible solution for the original problem can be reconstructed as
(x1 , x2 , x3 ) = (−1, 1, 0). A similar conversion can, of course, be executed for the optimal
solution.

Finally, we observe that in the new problem, terms involving x+ −


3 and x3 always appear in
the form of a pair, with opposite signs. This property has an interesting consequence. Recall
that all of the operations in the Simplex algorithm are row operations. This implies that
the above property will be preserved throughout the Simplex iterations. Now, suppose x+ 3
is a basic variable in a tableau. Then, since every column associated with a basic variable
must have its entries equal to 0 except for one that equals 1, it follows that the x−
3 -column

2
assumes the form
x−
3
0
0
..
.
0
−1
0
..
.
0

where the coefficient −1 is located at some row. Clearly, the variable associated with such
a column cannot be basic. Therefore, if x+ −
3 happens to be basic, then x3 must be nonbasic;
+
and conversely, if x−
3 happens to be basic, then x3 must be nonbasic. Of course, both of
these variables can be nonbasic at the same time. In summary, the conclusion is that, in
any basic feasible solution, at most one of the two values assigned to x+ −
3 and x3 can be
positive. Remarkably, this is in complete agreement with the (intended) interpretation of
x+ −
3 and x3 as the positive part and the negative part of the variable x3 .

You might also like