LS-OPT 5.1

The following features have been added:

1. Multilevel optimization. An LS-OPT solver type can be selected to allow the nesting of any LS-OPT task.
   
2. Parallel Feedforward Neural Networks. This feature allows the concurrent building of multiple networks and network ensemble components. FFNN building can also be done remotely, e.g. on a cluster. Job monitoring is provided in the GUI.
3. Significant enhancements have been made to histogram displays in the Viewer. Manual axis control is allowed while statistical quantities such as mean and standard deviation as well as constraints are depicted. Histogram types have been added.
   
4. Subregion-based sensitivity analysis is available using Sobol indices. Multiple subregions can be analyzed in the same run and stored for display. Global Sensitivity Analysis can now be activated from the GSA icon (as a post-processing function).
5. Design categories can be specified for user-selected simulation points. Name, color and type attributes can be chosen for each category display. Responses and histories are supported.
6. Excel is now supported as a solver type on Windows.
7. A new third party Finite Element solver is now supported. The support includes parameter (*PARAMETER) recognition using recursive include files during the problem setup phase.
8. De-activation of variables in iterative methods. The user can seamlessly deactivate variables at any stage of the iterative process. This is useful for variable screening.
9. Metamodel formulae for polynomials and Radial Basis Function Networks can be exported.
10. Multiple plots are allowed in optimization history displays. All the available entities such as variables, responses, etc. can be displayed on the same plot.
    
11. Differential Evolution was added as a global metamodel optimizer (unconstrained continuous problems only).
12. Responses and/or histories can be cloned (Stage dialog in GUI).