button or double click on the Dispersive Curves list in the Dispersive Model Fitting Configuration dialogue invoke Dispersive Curve dialogue for configuring (loading) ε’ (Eps’) and ε” (Eps’’) dispersion curves.6.2 Dispersive Curve
The button or double click on the Dispersive Curves list in the Dispersive Model Fitting Configuration dialogue invoke Dispersive Curve dialogue for configuring (loading) ε’ (Eps’) and ε” (Eps’’) dispersion curves.
The Name field contains the name of the dispersive curve.
Single-, dual-, and triple-pole dispersive medium model, being a composition of Debye, Drude or Lorentz dispersion models, can be chosen as a target for fitting procedure.
- invokes Load Dispersive Eps dialogue for loading appropriate dispersive curve
- invokes Results window to graphically preview the loaded ε’ (Eps’) and ε” (Eps’’) dispersive curves.
Process operation parameters of dispersion curves fitting procedure can be configured in the Preferences tab of Dispersive Curve dialogue.
Curve Fitting Parameters-Stop Criteria group contains the following options:
· Max nb of Simulations – indicates the number of runs after which the fitting process stops regardless of the obtained results.
· Stop when Goal Function Reaches – indicates the level of the satisfactory level of the goal function. The fitting process will stop when the goal function value drops below (Goal Function Target + Goal Function Tolerance), provided that the Stop when Goal Function Stabilises within Tolerance option is checked with Goal Function Tolerance defined. A typically used value of the Goal Function Target is zero.
· Stop when Goal Function Stabilises within Tolerance - When this option is OFF, QW-OptimiserPlus will not stop, when it reaches the Goal Function Target. It will continue searching for the best possible solution until Max nb of Simulations is completed. The Goal Function Tolerance defined influences the previous stop criterion. Additionally, it causes the optimisation process to stop when the Goal Function seems to have reached a local minimum. The optimisation will stop when the following three conditions are met: there are at least 2*N points in cache (where N denotes the number of optimisation variables); the best 2*N points are within variables tolerances from the single best point; and the goal function values at the best 2*N points are within Goal Function Tolerance from the single best point.
Curve Fitting-Initial Search group contains the following options:
· Step Size – indicates the initial area of search for the optimum. It is expressed as a fraction of the range of each of the variables. The value of 0.2 means that the solution will be initially searched with modification of each of the variables in the range of ±10%. Higher value enhances the chances of finding the global minimum but may slow down the optimisation process when one has a good starting point.
· Number of Random Searches – if equal to k>0, it instructs QW-OptimiserPlus to perform k global random searches. Deterministic optimisation is then started locally from the best available starting point.
· Number of Local Searches – if equal to k>0, it instructs QW‑OptimiserPlus to perform k initial steps exploring the vicinity of the specified starting point, without reference to cache. After these k steps, it compares the generated k points with cache, and re-starts further deterministic optimisation locally from the best available starting point.
Note that the default values of Initial Search parameters are 0.15; 0; 0. It is difficult to give very precise rules when those default values must or should be changed. This will depend on particular properties of the optimised project. Thus the user is encouraged to experiment with changing them in the case of difficult, slow-converging tasks, taking into account the hints provided above.