SCONE Optimizers

The goal of an Optimizer is to find the parameters for which an objective function is minimized (or maximized, depending on the type of objective). SCONE uses a shooting-based approach to optimization, in which a simulation is performed multiple times, and parameters are adjusted according the result of the objective function.

Optimization Algorithms

Currently, two types of Optimizers are available in SCONE: CmaOptimizer and CmaPoolOptimizer. Both are variants of Covariance Matrix Adaptation [Hansen 2006].

Stop Conditions

Optimizers will run continue to run until one of the following conditions has been met:

  • The user has aborted the optimization, either through the menu or by pressing the 'x' in the top right corner of the progress window.
  • The optimization has reached its maximum number of iterations / generations, set by `max_generations` (see CmaOptimizer).
  • The fitness has stopped improving, as set by the `min_progress` and `min_progress_samples` settings (see CmaOptimizer).
  • There is no difference between the fitness of individuals within one generation, so-called flat fitness.
  • by tgeijten