Explanation of the content
--------------------------
File: 151209-gecco2015algs-scmatls-1run.csv
The algorithms presented in http://dl.acm.org/citation.cfm?id=2754716 are run once on the 9720 instances from http://dl.acm.org/citation.cfm?id=2598249.
Author: Markus Wagner, wagner@acrocon.com
Columns B to M are the raw objective scores (we are maximising) of a single run.
The maximum of that set of experiments is recorded in O.
Based on O, the scores are translated to approximation ratios in Q to AB, which is why these are called "...apphere".
In AD, I listed the max values that we observed in the GECCO2015 paper.
Based on this and on O from the "here" experiments, the best ever observed by us is in AF "maxHereAndGECCO2015".
Based on AF, the approximation ratios are listed in AH to AS.
If you ask "Which are relevant?": I'd take AH to AS (to be maximised), since this represents "approximation ratio with respect to best ever seen by us".
Note that "approximation" here means that the results are scaled from the interval [worstAverageByAnAlgorithm, bestAverageByAnAlgorithm] into [0,1].
As MATLS did not produce results for some instances, its value is for these instances is -1.
The article in which all algorithms are described is the following:
Hayden Faulkner, Sergey Polyakovskiy, Tom Schultz, and Markus Wagner. 2015.
Approximate Approaches to the Traveling Thief Problem.
In Proceedings of the 2015 Annual Conference on Genetic and Evolutionary Computation (GECCO '15), Sara Silva (Ed.).
ACM, New York, NY, USA, 385-392.
DOI=http://dx.doi.org/10.1145/2739480.2754716