Can LS-OPT handle discrete variables?

Discrete Variables

To define a discrete variable in LS-OPT, the user has to specify a name, a starting value and a list of allowable values.

Mixed-discrete optimization

Discrete variables can be used together with continuous variables. This is called mixed-discrete optimization. The steps to compute a mixed-discrete optimum are

• Consider all the discrete variables to be continuous and optimize. This continuous optimum found is used as the starting design in the next phase.
• Discrete optimization is done considering only the discrete variables with the continuous variables frozen at the values found in the previous phase.
• Continuous optimization is done considering only the continuous variables and with the discrete variables frozen at the values found in the previous phase.

Example: Discrete optimization

Sampling at discrete points

A flag is provides to select the sampling points at specified discrete values of the variables if discrete variables are specified. Discrete sampling will also handle discrete-continuous problems correctly. In the GUI, a switch box is provided for each discrete variable in the variables panel to define discrete sampling.

If possible, a continuous sampling is recommended, because it usually leads to a better distribution of the points within the design space and hence to a better metamodel quality.

String Variables

As a special type of discrete variables, string variables may be defined in LS-OPT.

To define a string variable in LS-OPT, the user has to specify a name, a starting variable and a list of allowable values.