Data sets and figures for the paper "Relating Training Instances to Automatic Design of Algorithms for Bin Packing via Features"; Alexander E.I. Brownlee, John R. Woodward, Nadarajen Veerapen; Companion Proceedings of GECCO 2018, Kyoto Japan.

figures/ is the full set of figures for the instance sets and footprints

stirling-instances/ contains the generated instances mentioned in the paper. Dir names are stir_generated_binCapacity_itemSizeLowerBound_itemSizeUpperBound. Filenames are stir_binCapacity_numberOfItems_itemSizeLowerBound_itemSizeUpperBound Format of files are the same as the Falkenauer benchmarks: 
  line1:number of items
  line2:volume of bins
  any addition lines: item size
  
data-wide/ contains the results of GP runs. There are three types of files in here:
 - waescher_instances.txt is just a list of the instance files in a given set (Waescher in this case)
 - waescher_performance_0.txt is the output from a GP run for the named instance set (Waescher here). Line 1 is the best GP program found. The remainder is tab-separated data for each bin packing instance in the set; instanceSet	instanceName(usually automatically generated)	numberOfBins(required when using the GP evolved packing policy)
 - waescher_29_on_stir_generated_150_76_100_performance.txt is the result when running the policy that was trained on waescher (run 29), on the instances in the stir_generated_150_76_100 set. content is same as line above.
 
metricsAndRawBinCountsAll-allbp-trained-footprints - comma separated file with all data used for the analysis. Each row is the data for bin bin packing instance. Columns are:
 - format: the file format
 - set: the instance set the instance belongs to
 - id: a unique ID
 - file: filename containing the instance
 - numberInFile: some of the files have more than one instance. This is the index within the file.
 - isVolumesAreInts-ItemListCompressionRatio: the static features for the instance
 - ScaledFit0_raw-ScaledFit10_raw: the performance features for the instance (number of bins required by the scaled fit policies)
 - ScaledFit0_rank-ScaledFit10_dist: scaled variants of the performance features (unused in the paper)
 - 2cbp_0-stir_generated_150_76_100_29: number of bins required for this instance when using the policy trained on the named set, in GP run number x
 - 2cbp_footprintCount-stir_generated_150_76_100_footprintCount: number of times this instances appears in the footprint of policies trained on the names instance set