Iteration Information

The estimate is computed by iterative search for a model that gives the minimal prediction error variance, when applied to the working data. The Parametric Model and the Process Model dialogs provide online information about the progress of this search.

The number of the current iteration is shown after 'Iteration #' together with the current Fit (Loss function = the sum of squares of output prediction errors normalized by the number of data; for multi-output models, it is the determinant of the covariance matrix of the prediction errors).

The improvement in fit compared to the previous iteration is also shown.

If you check the box 'Trace', a full trace of loss functions, parameter values, and update information is given in the MATLAB command window.

Press 'Stop iterations' to stop the search and save the result after the current iteration has been completed. Press 'Continue' to continue the iterations from the current model.

The iterative search is controlled by a number of design variables that can be accessed by pressing the Iteration options... button.

Help topics.

(file iduiiter.htm)