SymbolicRegressionSolutionsAnalyzer should also add training and test R² values to the scope so that they can be read by other analyzers

[gkronber]

It is currently not possible to track test R² (or other metrics for test quality) of symbolic regression solutions with the analyzers in HeuristicLab.Analysis because this value is not stored in the scope.

[gkronber]

r13582: added parameters to SymbolicRegressionSolutionsAnalyzer to write R² (training & test) for all solutions to the scope

[mkommend]

Review comments:

• What is the point of this ticket (hard to judge without a description)? It is hard for other analyzers to read the R² values, because they would have to know the solution names, because these are used as variable names.
• Why are there always null checks before accessing parameters? Aren't these obsolete, because the parameter is never null?

[gkronber]

Replying to mkommend:

Review comments:

• What is the point of this ticket (hard to judge without a description)? It is hard for other analyzers to read the R² values, because they would have to know the solution names, because these are used as variable names.

It is currently not possible to track test R² (or other metrics for test quality) of symbolic regression solutions with the analyzers in HeuristicLab.Analysis because this value is not stored in the scope. AFAIK, this was the reason why this analyzer was implemented originally. The analyzer also creates a chart of training and test R². This can also be achieved using the standard analyzers when the test quality values are stored in the scope.

[gkronber]

r13660: removed null checks

[mkommend]

Reviewed r13582 and r13660.

[mkommend]

r13950: Merged r13582 and r13660 into stable.