Context Navigation

#1823 closed feature request (done)

Symbolic data analysis analyzer that collects the Pareto-optimal solutions regarding complexity and accuracy

Reported by:	gkronber	Owned by:	gkronber
Priority:	medium	Milestone:	HeuristicLab 3.3.7
Component:	Problems.DataAnalysis.Symbolic	Version:	3.3.7
Keywords:		Cc:

Description

Change History (17)

comment:1 Changed 13 years ago by gkronber

Status changed from new to accepted

comment:2 Changed 13 years ago by gkronber

r7726:

added analyzer that calculates the complexity of symbolic data analysis trees (weighted symbols)
added analyzer that collects the Pareto-optimal solutions regarding complexity and accuracy

comment:3 Changed 13 years ago by gkronber

r7734: implemented analyzers to collect Pareto-best solutions on validation and fitness calculation partitions for regression and classification.

comment:4 Changed 13 years ago by gkronber

r7739: set complexity analyzer for symbolic data analysis solutions to disabled by default

comment:5 Changed 12 years ago by swinkler

When using a SingleObjectiveTrainingParetoBestSolutionAnalyzer (in r7967) an exception occurs:

An exception was thrown by the operator "SymbolicClassificationSingleObjectiveTrainingParetoBestSolutionAnalyzer": Index was out of range. Must be non-negative and less than the size of the collection. Parameter name: index

Last edited 12 years ago by swinkler (previous) (diff)

comment:6 Changed 12 years ago by gkronber

r8126:

fixed the bug in the Pareto-best solution analyzers for symbolic data analysis (as reported in comment:5).
fixed a minor bug in the calculation of classification thresholds
made a sneaky change in SymbolicExpressionTree

comment:7 Changed 12 years ago by gkronber

Owner changed from gkronber to mkommend
Status changed from accepted to reviewing

comment:8 Changed 12 years ago by gkronber

r8130: added AfterDeserialization hook to add the new analyzers when loading old problems.

comment:9 Changed 12 years ago by mkommend

reviewing comments:

I would move the problem data parameter one level up to SymbolicDataAnalysisSingleObjectiveTrainingParetoBestSolutionAnalyzer instead of adding it for both classification and regression parameters.
The same holds for the interpreter parameter.
IMHO the solution in the ComplexityAnalyzer to get the default weights is not optimal, because it has to be updated for every new symbol, it doesn't work if a symbol name is changed in the grammar editor and the dictionary is created for every execution of the analyzer.