MPL for Windows Manual
MPL for Windows Manual
 |
 |
 |
You can change the simplex file options for XPRESS by choosing XPRESS parameters from the Options menu and then pressing the Simplex tab. This will display the simplex options for XPRESS in the dialog box as shown below:
Figure 4.53: The XPRESS Simplex Options Dialog Box
The algorithm to used by the optimizer to solve the problem can be specified here. The default value is Automatic selection. Following are the four possible selections:
Automatic selection Dual simplex for LPs and IPs; Newton barrier for QP. Primal algorithm Use primal simplex algorithm. Dual algorithm Use dual simplex algorithm. Newton barrier Use Newton barrier algorithm.
The type and severity of the crash to be performed for primal simplex can be choosing one of the following options:
Off Turns off all crash procedures. Singletons / One pass For singletons only, one pass.
Multi pass / Slacks
Multi pass / Slack at e If primal, alternate ways of using objective coefficients; else, if dual default starting basis (default).
The Pricing option for primal simplex has the following possible selections:
Automatic selection The pricing strategy is to be decided automatically (default). Partial pricing Partial pricing is to be used. Devex pricing Devex pricing is to be used.
Infeasbility sets controls the number of irreducible infeasible sets to be found. It has the following possible selections:
Search for all (default). Search for none. Search for first n.
The Use perturbation is used to select whether to use perturbation or not, prior to optimization. You can choose either Automatic perturbation or Manual peturbation; if you choose Manual, you can set the perturbation value (default is 0, i.e., no perturbation).
The Initial Solution allows you to choose between the Phase I/Phase II method or the Big M method (the default) to obtain an initial feasible solution.
Phase I / Phase II For PhaseI/PhaseII. Big M Big M method to be used. You can set the infeasibility Penalty which has the default of 1024 (default).
It is always worth striving to create a well-scaled matrix during the formulation stage. This is not always easy to do, and so automatic scaling can be applied to improve the numerical stability, if the range of coefficient values of a matrix is very large. The following are the five available scaling techniques
Row scaling. Column scaling. Row scaling again. Maximin scaling. Curtis-Reid scaling.
Furthermore, you can choose between scaling by geometric mean or by maximum element. The default is Row and Column scaling, by maximum element.
Scaling of integer entities is not supported, though XPRESS will scale the continuous variables in a MIP problem. This means that you should be careful about the scale of the integer entities when formulating MIP problems.