19.03.2018 Views

linear-guest

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

3.3 Dantzig’s Algorithm 77<br />

⎛<br />

⎜<br />

⎝<br />

variables (incl. slack and artificial)<br />

{ }} { objective<br />

{}}{<br />

⎞<br />

⎟<br />

⎠<br />

↑<br />

objective value<br />

← constraint equations<br />

← objective equation<br />

Figure 3.1: Arranging the information of an optimization problem in an<br />

augmented matrix.<br />

Example 41 (Performing EROs)<br />

We scan the last row, and notice the (most negative) coefficient −4. Naïvely you<br />

might think that this is good because this multiplies the positive variable w and only<br />

helps the objective function f = 4w + · · · . However, what this actually means is<br />

that the variable w will be positive and thus determined by the constraints. Therefore<br />

we want to remove it from the objective function. We can zero out this entry by<br />

performing a row operation. For that, either of the first two rows could be used.<br />

To decide which, we remember that we still have to solve solve the constraints for<br />

variables that are positive. Hence we should try to keep the first two entries in the<br />

last column positive. Hence we choose the row which will add the smallest constant<br />

to f when we zero out the −4: Look at the last column (where the values of the<br />

constraints are stored). We see that adding four times the first row to the last row<br />

would zero out the −4 entry but add 20 to f, while adding two times the second row<br />

to the last row would also zero out the −4 but only add 12 to f. (You can follow this<br />

by watching what happens to the last entry in the last row.) So we perform the latter<br />

row operation and obtain the following:<br />

⎛<br />

⎞<br />

⎜<br />

⎝<br />

1 1 1 1 0 5<br />

1 2 3 2 0 6<br />

−1 7 7 0 1 12<br />

⎟<br />

⎠<br />

c 1 = 5<br />

c 2 = 6<br />

f = 12 + x − 7y − 7z .<br />

We do not want to undo any of our good work when we perform further row operations,<br />

so now we use the second row to zero out all other entries in the fourth column. This<br />

is achieved by subtracting half the second row from the first:<br />

⎛<br />

⎜<br />

⎝<br />

1<br />

2<br />

0 − 1 2<br />

0 0 2<br />

1 2 3 2 0 6<br />

−1 7 7 0 1 12<br />

⎞<br />

⎟<br />

⎠<br />

c 1 − 1 2 c 2 = 2<br />

c 2 = 6<br />

f = 12 + x − 7y − 7z .<br />

77

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!