Two Weighting Local Search for Minimum Vertex Cover

Authors: Shaowei Cai, Jinkun Lin, Kaile Su

AAAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments show that Tw MVC outperforms Nu MVC on the standard benchmarks namely DIMACS and BHOSLIB. To the best of our knowledge, Tw MVC is the first Min VC algorithm that attains the best known solution for all instances in both benchmarks.
Researcher Affiliation Academia 1State Key Laboratory of Computer Science, Institute of Software, Chinese Academy of Sciences, Beijing, China 2Key Laboratory of High Confidence Software Technologies, Peking University, Beijing, China 3Institute for Integrated and Intelligent Systems, Griffith University, Brisbane, Australia
Pseudocode Yes Algorithm 1: A Local Search Framework of Min VC; Algorithm 2: Tw MVC; Algorithm 3: choose Add V ertex(e)
Open Source Code Yes Tw MVC is programmed in C++, on the basis of source codes of Nu MVC6. Available from http://lcs.ios.ac.cn/~caisw/MVC.html
Open Datasets Yes The DIMACS benchmark is taken from the Second DIMACS Implementation Challenge for the Maximum Clique problem (1992-1993)3... The BHOSLIB4 (Benchmarks with Hidden Optimum Solutions) instances were generated in the phase transition area according to model RB (Xu et al. 2007)... The web link networks5 were generated from web pages online.
Dataset Splits No The paper does not specify training/test/validation dataset splits, as it concerns combinatorial optimization algorithms evaluated on entire benchmark instances rather than machine learning models trained on split data.
Hardware Specification Yes All experiments are carried out on a machine under Linux, using an Intel(R) Core(TM) 2.8 GHz CPU and 4 GB RAM.
Software Dependencies Yes Both algorithms are complied by g++ (version 4.4.5) with the -O2 option.
Experiment Setup Yes The parameters in Tw MVC are tuned manually based on experiments, and are set as follows. Setting β and δ: ... For all instances: β = 0.8 and δ = 100000. Setting γ: ... Finally, γ is set to |E|/7 for most instances, except for C instances (where γ = |E|/15), MANN instances (where γ = |E|/3.3) and web instances (where γ = |E|/10). Although the above setting can already yield good performance on brock instances, we observe a much better setting for brock instances is: β = 1, delta = 10000 and γ = 1500, which is adopted in our experiment.