Analysis | Function Optimization

New Menu Item

Following the same logic as the one used for the reorganization of the Target Optimization tools, we have created this new menu item to regroup the three new tools dedicated to the optimization of Function Nodes.

New Feature: Optimization Score to Minimize

As there is no direct way to get the theoretical range of Function Nodes' values, there are no Minimization/Maximization options as those in the Target Optimization tools.

A target value has thus to be set for the to-be optimized Function Node in order to compute the score (the default target value is the current value of the node).

The score is defined as follows:

where is the target value and the value of the Function Node of the evaluated solution.

When Resources are defined, a target value has also to be set (as BayesiaLab does not have the range). The default value is the sum of all the Function nodes associated with the predefined class Resource.

The score is defined as follows:

where is the target value and the value of the Resources of the evaluated solution.

The Weight is used to compute the overall optimization score:

New Feature: Predefined Class Opt_Hard_Evidence

The Search Method defined in the three optimization tools applies to all the variables. However, sometimes the optimization requires a mix between Mean/Value Variations and Hard Evidence (e.g. for variables with states that cannot be ordered). In this context, version 7.0 features a new Predefined Class called Opt_Hard_Evidence. Using this class allows setting the search method to Mean/Value Variations, while still being able to optimized the variables associated with this new class with Hard Evidence.