Data Set to Accompany "Quantifying the Impact of Parameter Tuning on Nature-Inspired Algorithms"

Performance data gathered in the experiments described in 'Quantifying the Impact of Parameter Tuning on Nature-Inspired Algorithms' (Crossley, Nisbet and Amos (2013)).

 

'Raw' performance data - that is to say, the best fitness from all runs performed - are available in the individual sheets labelled Untuned/Tuned , where performance data is split across 5 sheets - one for each characteristic.

 

Summary.xlsx contains a summary of all the data - average performance/standard deviation for each characteristic value, for each algorithm.

 

Parameters.xlsx contains a summary of the parameters used, both the 'default' (untuned) and tuned parameters selected by the F-Racing process.

 

The accompanying paper has been submitted to the European Conference on Artificial Life 2013 (ECAL13).