Experiments with 3-opt local search
ACO and CGS algorithms work better if at each iteration the solutions are improved using a local search procedure. In a series of experiments, we have compared the performance of ACO and CGS algorithms combined with 3-opt local search. These experiments are described in this paper, presented at PPSN 2010. In the results archive, you will find two main folders: experiments and tuning.
The tuning folder contains subfolders corresponding to each of the algorithms considered: acs, cgs, cgsf and mmas. For each algorithm, the result files of the tuning phase are provided, along with the configuration files specifying the lists of values used for generating the competing parameter settings.
The experiments folder also contains a subfolder for each of the algorithms. The experimental results are available in the files named end.result.txt. A compact report comprising the percentage deviation from optimum for the best and mean results, as well as the standard deviations of the results, is given in the files named compact.txt.