Constrained Quadratic Optimisation: 4.
Updating the tableau
[this page | pdf | references | back links | custom searches]
Abstract and Contents
4. Updating the
We update the tableau iteratively (probably only up to some
upper limit of number of iterations, in case there is an error in the
computation), and we stop when there is no change to the tableau at a given
To do this we need to identify which variable ought ideally
to enter the feasible solution, i.e. to become a ‘basic’ variable, and which
variable it should replace, i.e. which one is ceasing to be a basic variable.
We need to do these simultaneously, since the constraints mean that only
certain combinations of variables can enter and leave at the same time.
This may be done by identifying the largest positive value
of the ObjectiveRow for a column (variable) which is not currently basic (but
only if either the current corresponding variable is non-basic, so that the
joint constraint of the form value is still
satisfied, or if the corresponding variable is basic then the two can be
swapped over and still improve the objective function) and if there is
another basic variable, which if removed from the feasible set at the same time
improves the objective function.
As long as we identify a one entering basic variable
(column) and one exiting basic variable (column) as above, we pivot the Tableau,
ObjectiveRow and SolutionColumn around their intersection and we
update the BasicRow and BasicColumn accordingly.
As mentioned above, sometimes we need to replace zeros with
very small positive numbers to avoid the tableau becoming degenerate.
Contents | Prev | Next