Class CMD_optimo

All Implemented Interfaces:
NBComponent, NBComponentAdvisors, NBComponentEvents, NBComponentMetrics, NBComponentProps, NBComponentServices, NBComponentTimeline, NBProviderSearch, NBTokenWords, NBLabeledElement, AutoCloseable, BiFunction<NBBufferedContainer,NBCommandParams,Object>

public class CMD_optimo extends NBBaseCommand
If the initial trust radius is not large enough with respect to the stopping trust radius, then the search will stop ascending the approximate manifold too early. This seems to be because the trust radius is forced to diminish by some limiting condition.

Explanation of parameters:

  1. numberOfInterpolationPoints - limited by the dimensionality of the parameter vector. M.J.D. Powell recommends this be 2n+1, where n is the number of parameters in the search vector.
  2. initialTrustRegionRadius - Initial trust region radius, affects how well aggressively BOBYQA explores the parameter space. Too small and it may not explore enough to find a gradient if it starts in a flatter part of the manifold. Too large, and the feedback loop seems to run dry quickly due to lack of detail.
  3. stoppingTrustRegionRadius - Stopping trust region radius. This is the space which contains interpolation points, and can be trusted to contain a reasonably representative character of the overall manifold