Closed HeuristicLab-Trac-Bot closed 9 years ago
r8739 removed static method calls to similarity calculators
r8797 added a comparison factor parameter to the crossover analyzer so that it can be defined what a successful crossover is
r8866 further improvements for the crossover performance analyzer
- improved unwanted mutation analyzer
- added test functions crossover performance analyzer
r9045 added mutation performance analyzer for test functions
r9053 added a dialog for adding operators as successors to other operators
r9054 added an analyzer for comparing the current solution to the rest of the population
r9061 added average of first, last, upper and lower 25% to statistical tabular view
r9150 fixed successor configuration and added successful mutation chart
r9192 RunCollectionStatisticalTabularView now respects RunCollectionFilters
- fixed a cloning bug
- fixed forgotten event handler deregistration
- added a run collection view for analyzing parameter influences
r9198 added calculation of the overall parameter influence
r9252 automatically calculate statistical values for each run
r9317 started working on a statistical testing run collection view
r9318 added test for normality and pairwise statistical testing
r9320 some more improvements for the statistical testing view
- fixed a bug in the ResultCorrelationView
- added sample size determination by estimating means, Cohen's D and Hedges' G
r9329 calculate Cohen's D and Hedges' G for pairwise comparison and updated documentation
r9336 added ttest and sample size determination for the ttest
r9354 removed views that went into the statistical hypothesis testing branch (#2031)
r9361 adapted create experiment dialog so that it can generate random configurations
r9364 added repetitions for random parameter configurations
r9412 added GA with evaluated solutions as a stopping criteria
r9412: I'm not sure it is necessary to add a GA variant with evaluated solutions as stopping criterion (this is almost the same as pop-size * generations)
Replying to [comment:98 gkronber]:
r9412: I'm not sure it is necessary to add a GA variant with evaluated solutions as stopping criterion (this is almost the same as pop-size * generations) Yes, that's true, but it makes creating experiments simpler as I can fix the evaluated solutions and just vary the population size. Generations are not varied and are then simply a result. Because I want to define ranges for parameters and then draw random samples it would be more complicated to check if a drawn configuration has the correct amount of evaluated solutions (or only calculated configurations where this is the case).
- changed duplicate solutions analyzer to also track avg. diversity as the standard diversity analyzer needs too much memory
- fixed project references of eslGA
Issue migrated from trac ticket # 1886
milestone: HeuristicLab 3.3.x Backlog | component: Analysis | priority: medium | resolution: obsolete
2012-06-29 13:15:07: @Shabbafru created the issue