SCCWalk: An efficient local search algorithm and its improvements for maximum weight clique problem

Artif. Intell., pp. 1032302020.

Cited by: 1|Bibtex|Views47|Links
EI
Keywords:
winner determination problemexact algorithmmaximum weight clique problemlarge graphmaximum clique problemMore(13+)
Weibo:
Experimental results on massive graphs show that the best from multiple selection heuristic and reduction rules significantly improve the performance of the algorithms on massive real-world graphs and large-scaled

Abstract:

Abstract The maximum weight clique problem (MWCP) is an important generalization of the maximum clique problem with wide applications. In this study, we develop two efficient local search algorithms for MWCP, namely SCCWalk and SCCWalk4L, where SCCWalk4L is improved from SCCWalk for large graphs. There are two main ideas in SCCWalk, inc...More

Code:

Data:

0
Introduction
  • The maximum clique problem (MCP) consists of finding a clique with the maximum number of vertices.
  • An important generalization of MCP is the maximum weight clique problem (MWCP), in which each vertex is associated with a non-negative integer, and the goal is to find a clique with the largest total weight.
  • A novel formulation of selecting object region candidates simultaneously in all frames is proposed, to find the maximum weight cliques in a weighted region graph [6].
  • Wu and Hao [8] study the winner determination problem (WDP) in combinatorial auctions by recasting this problem into MWCP
Highlights
  • An important generalization of maximum clique problem (MCP) is the maximum weight clique problem (MWCP), in which each vertex is associated with a non-negative integer, and the goal is to find a clique with the largest total weight
  • To further demonstrate the performance of SCCWalk on winner determination problem (WDP) benchmarks, we tested SCCWalk on 500 small instances with up to 1500 items and 1500 bids, which can be divided into five groups
  • We developed two local search algorithms for the maximum weight clique problem (MWCP)
  • SCCWalk4L used two reduction rules with some limitations to reduce the size of the MWCP problem
  • Experimental results on massive graphs show that the best from multiple selection (BMS) heuristic and reduction rules significantly improve the performance of the algorithms on massive real-world graphs and large-scaled
Results
  • Experimental results with five competitors on the

    DIMACS benchmark. Experimental results on the

    DIMACS are shown in Table 3.
  • The results shown in Table 6 illustrate that SCCWalk has a better performance than other competitors in these instances except for Vowel, where WLMC takes a shorter time to find the same solution quality.
  • The superiority of SCCWalk on the classical benchmarks is clearly illustrated by Fig. 2, which summarizes the run time distributions of the MWCP algorithms when both SCCWalk and one competitor find the maximum solution values
Conclusion
  • The authors developed two local search algorithms for the maximum weight clique problem (MWCP).
  • The abovementioned heuristics were used to develop a local search algorithm named SCCWalk.
  • SCCWalk4L used two reduction rules with some limitations to reduce the size of the MWCP problem.
  • Experimental results on massive graphs show that the BMS heuristic and reduction rules significantly improve the performance of the algorithms on massive real-world graphs and large-scaled frb18-1-2.2-10-0.19748 frb18-2-2.2-10-0.19748 frb18-3-2.2-10-0.19748 frb18-4-2.2-10-0.19748 frb18-5-2.2-10-0.19748 frb19-1-2.2-10-0.19748 frb19-2-2.2-10-0.19748 frb19-3-2.2-10-0.19748 frb19-4-2.2-10-0.19748 frb19-5-2.2-10-0.19748 frb20-1-2.2-10-0.19748 frb20-2-2.2-10-0.19748 frb20-3-2.2-10-0.19748 frb20-4-2.2-10-0.19748 frb20-5-2.2-10-0.19748 frb21-1-2.2-10-0.19748 frb21-2-2.2-10-0.19748 frb21-3-2.2-10-0.19748 frb21-4-2.2-10-0.19748 frb21-5-2.2-10-0.19748
Summary
  • Introduction:

    The maximum clique problem (MCP) consists of finding a clique with the maximum number of vertices.
  • An important generalization of MCP is the maximum weight clique problem (MWCP), in which each vertex is associated with a non-negative integer, and the goal is to find a clique with the largest total weight.
  • A novel formulation of selecting object region candidates simultaneously in all frames is proposed, to find the maximum weight cliques in a weighted region graph [6].
  • Wu and Hao [8] study the winner determination problem (WDP) in combinatorial auctions by recasting this problem into MWCP
  • Results:

    Experimental results with five competitors on the

    DIMACS benchmark. Experimental results on the

    DIMACS are shown in Table 3.
  • The results shown in Table 6 illustrate that SCCWalk has a better performance than other competitors in these instances except for Vowel, where WLMC takes a shorter time to find the same solution quality.
  • The superiority of SCCWalk on the classical benchmarks is clearly illustrated by Fig. 2, which summarizes the run time distributions of the MWCP algorithms when both SCCWalk and one competitor find the maximum solution values
  • Conclusion:

    The authors developed two local search algorithms for the maximum weight clique problem (MWCP).
  • The abovementioned heuristics were used to develop a local search algorithm named SCCWalk.
  • SCCWalk4L used two reduction rules with some limitations to reduce the size of the MWCP problem.
  • Experimental results on massive graphs show that the BMS heuristic and reduction rules significantly improve the performance of the algorithms on massive real-world graphs and large-scaled frb18-1-2.2-10-0.19748 frb18-2-2.2-10-0.19748 frb18-3-2.2-10-0.19748 frb18-4-2.2-10-0.19748 frb18-5-2.2-10-0.19748 frb19-1-2.2-10-0.19748 frb19-2-2.2-10-0.19748 frb19-3-2.2-10-0.19748 frb19-4-2.2-10-0.19748 frb19-5-2.2-10-0.19748 frb20-1-2.2-10-0.19748 frb20-2-2.2-10-0.19748 frb20-3-2.2-10-0.19748 frb20-4-2.2-10-0.19748 frb20-5-2.2-10-0.19748 frb21-1-2.2-10-0.19748 frb21-2-2.2-10-0.19748 frb21-3-2.2-10-0.19748 frb21-4-2.2-10-0.19748 frb21-5-2.2-10-0.19748
Tables
  • Table1: Information of the DGCA benchmark
  • Table2: The parameter settings used in SCCWalk on DIMACS, BHOSLIB, WDP and DGCA benchmarks
  • Table3: Experimental results of ReTS1, ReTS2, FastWClq, RRWL, WLMC, and SCCWalk on the DIMACS benchmark
  • Table4: Experimental results of ReTS1, ReTS2, FastWClq, RRWL, WLMC and SCCWalk on the BHOSLIB benchmark
  • Table5: Experimental results of ReTS1, ReTS2, FastWClq, RRWL, WLMC and SCCWalk on the WDP benchmark
  • Table6: Experimental results of ReTS1, ReTS2, FastWClq, RRWL, WLMC and SCCWalk on the DGCA benchmark
  • Table7: Experimental results of CCWalk and SCCWalk on the classical benchmarks. A positive δmax or δavg indicates SCCWalk finds better quality clique than CCWalk
  • Table8: Comparative results for SCCWalk and five competitors on the classical benchmarks
  • Table9: Information of large-scaled FRB instances
  • Table10: Experimental results of ReTS1, ReTS2, FastWClq, RRWL, WLMC and SCCWalk4L on the real-world massive graphs I
  • Table11: Experimental results of ReTS1, ReTS2, FastWClq, RRWL, WLMC and SCCWalk4L on the real-world massive graphs II
  • Table12: Among the five competitors, FastWClq, RRWL, WLMC, and ReTS2 perform badly, so we focus on the comparison between ReTS1 and SCCWalk4L. Overall, SCCWalk4L finds better solutions than ReTS1, on these graphs with massive edges. We observe that ReTS1 can not reach maximum weight clique for 13 graphs, while SCCWalk4L finds maximum weight cliques for all 20 graphs. Experimental results of ReTS1, ReTS2, FastWClq, RRWL, WLMC and SCCWalk4L on the large-scaled FRB instances
  • Table13: Experimental results of SCCWalk and SCCWalk4L on representative classical and massive graphs
  • Table14: Comparative results for SCCWalk4L and five competitors on the massive graphs
Download tables as Excel
Related work
  • As is well known, the decision version of MCP is one of Karp’s prominent 21 NP-complete combinatorial problems [9]. Both MCP and MWCP have been proved to be NP-hard, and the current best-known approximation algorithms can only achieve an approximate ratio of O(n(log log n)2/(log n)3) [10]. Thus, a considerable amount of effort has been devoted to finding a “good” clique within a reasonable time. Up to now, there are mainly two types of algorithms for MCP and MWCP, i.e., exact and heuristic algorithms.
  • Related works on exact algorithms for

    MCP and MWCP A number of exact algorithms have been proposed to solve MCP and MWCP. A classic branch and bound algorithm was

    MCQ [11], which used a vertex ordering heuristic for independent set partition. The MCQ algorithm was further improved by computing the degrees of vertices dynamically, resulting in the MaxCliqueDyn algorithm [12]. Li and Quan [13] proposed a new encoding from MCP into MaxSAT, and then applied MaxSAT reasoning to improve the upper bound. San et al [14] presented the BBMCSP algorithm, which was based on a leading bit-parallel non-sparse solver, and used a novel sparse encoding for the adjacency matrix. MaxCLQ [15] combined an efficient preprocessing procedure and incremental MaxSAT reasoning in a branch and bound scheme. Recently, MoMC [16] combined a dynamic vertex ordering heuristic and a static vertex ordering heuristic during the search.
  • Related works on heuristic algorithms for

    MCP and MWCP Although exact algorithms can guarantee the optimality of their solutions, they may fail to solve hard instances of large scale. For solving large-sized instances, a popular approach is local search, which can find an approximate solution within a reasonable time. There are numerous local search algorithms for MCP [24,25,26,27,28,29]. Among these, DLS [25] was a milestone algorithm that used vertex penalties, and dynamically adjusted them during the search. DLS was further improved into a two-phase algorithm called phased local search (PLS) [26]. Wu and Hao proposed a tabu search algorithm with a k-fixed penalty strategy [28]. Benlic and Hao proposed a breakout local search algorithm for MCP [29]. MCP is closely related to minimum vertex cover (MinVC) and maximum independent set problems, and algorithms for these two problems can be directly used to solve MCP.
Funding
  • This work is supported by the Fundamental Research Funds for the Central Universities 2412018QD022 and 2412018ZD017, NSFC (under grant Nos. 61806050, 61972063, 61976050, 61972384)
  • Shaowei Cai was supported by Youth Innovation Promotion Association, Chinese Academy of Sciences (No 2017150)
Reference
  • D. Ballard, C. Brown, Computer Vision, Englewood Cliffs, Prentice Hall, 1982.
    Google ScholarFindings
  • B. Balasundaram, S. Butenko, Graph domination, coloring and cliques in telecommunications, in: Handbook of Optimization in Telecommunications, Springer, 2006, pp. 865–890.
    Google ScholarLocate open access versionFindings
  • M. Gomez Ravetti, P. Moscato, Identification of a 5-protein biomarker molecular signature for predicting Alzheimer’s disease, PLoS ONE 3 (9) (2008)
    Google ScholarLocate open access versionFindings
  • J. Pattillo, N. Youssef, S. Butenko, Clique relaxation models in social network analysis, in: Handbook of Optimization in Complex Networks, Springer, 2012, pp. 143–162.
    Google ScholarLocate open access versionFindings
  • F. Mascia, E. Cilia, M. Brunato, A. Passerini, Predicting structural and functional sites in proteins by searching for maximum-weight cliques, in: Twenty-Fourth AAAI Conference on Artificial Intelligence, AAAI 2010, Atlanta, Georgia, USA, July 2010, pp. 1274–1279.
    Google ScholarLocate open access versionFindings
  • T. Ma, L.J. Latecki, Maximum weight cliques with mutex constraints for video object segmentation, in: 2012 IEEE Conference on Computer Vision and Pattern Recognition, CVPR, IEEE, 2012, pp. 670–677.
    Google ScholarLocate open access versionFindings
  • H. Zhian, M. Sabaei, N.T. Javan, O. Tavallaie, Increasing coding opportunities using maximum-weight clique, in: Computer Science and Electronic Engineering Conference, 2013, pp. 168–173.
    Google ScholarLocate open access versionFindings
  • Q. Wu, J.-K. Hao, Solving the winner determination problem via a weighted maximum clique heuristic, Expert Syst. Appl. 42 (1) (2015) 355–365.
    Google ScholarLocate open access versionFindings
  • R. Karp, Reducibility among combinatorial problems, in: Complexity of Computer Computations, 1972, pp. 85–103.
    Google ScholarFindings
  • U. Feige, Approximating maximum clique by removing subgraphs, SIAM J. Discrete Math. 18 (2) (2004) 219–225.
    Google ScholarLocate open access versionFindings
  • E. Tomita, T. Seki, An efficient branch-and-bound algorithm for finding a maximum clique, in: International Conference on Discrete Mathematics and Theoretical Computer Science, Springer, 2003, pp. 278–289.
    Google ScholarLocate open access versionFindings
  • J. Konc, D. Janezic, An improved branch and bound algorithm for the maximum clique problem, Commun. Math. Comput. Chem. 58 (2007) 569–590.
    Google ScholarLocate open access versionFindings
  • C.M. Li, Z. Quan, An efficient branch-and-bound algorithm based on MaxSAT for the maximum clique problem, in: AAAI, vol. 10, 2010, pp. 128–133.
    Google ScholarFindings
  • P. San Segundo, A. Lopez, P.M. Pardalos, A new exact maximum clique algorithm for large and massive sparse graphs, Comput. Oper. Res. 66 (2016)
    Google ScholarLocate open access versionFindings
  • H. Jiang, C.M. Li, F. Manyà, Combining efficient preprocessing and incremental MaxSAT reasoning for MaxClique in large graphs, in: ECAI, 2016, pp. 939–947.
    Google ScholarFindings
  • C.-M. Li, H. Jiang, F. Manyà, On minimization of the number of branches in branch-and-bound algorithms for the maximum clique problem, Comput. Oper. Res. 84 (2017) 1–15.
    Google ScholarLocate open access versionFindings
  • L. Babel, A fast algorithm for the maximum weight clique problem, Computing 52 (1) (1994) 31–38.
    Google ScholarLocate open access versionFindings
  • P.R. Östergård, A new algorithm for the maximum-weight clique problem, Nord. J. Comput. 8 (4) (2001) 424–436.
    Google ScholarLocate open access versionFindings
  • K. Yamaguchi, S. Masuda, A new exact algorithm for the maximum weight clique problem, in: ITC-CSCC: 2008, 2008, pp. 317–320.
    Google ScholarFindings
  • S. Shimizu, K. Yamaguchi, T. Saitoh, S. Masuda, Optimal table method for finding the maximum weight clique, in: Proc. of the 13th International Conference on Applied Computer Science, 2013, pp. 84–90.
    Google ScholarLocate open access versionFindings
  • Z. Fang, C.-M. Li, K. Xu, An exact algorithm based on MaxSAT reasoning for the maximum weight clique problem, J. Artif. Intell. Res. 55 (2016) 799–833.
    Google ScholarLocate open access versionFindings
  • H. Jiang, C.-M. Li, F. Manya, An exact algorithm for the maximum weight clique problem in large graphs, in: AAAI, 2017, pp. 830–838.
    Google ScholarFindings
  • H. Jiang, C.-M. Li, Y. Liu, F. Manya, A two-stage MaxSAT reasoning approach for the maximum weight clique problem, in: Proceedings of the Thirty-Second AAAI Conference on Artificial Intelligence, 2018, pp. 1338–1346.
    Google ScholarLocate open access versionFindings
  • A. Singh, A.K. Gupta, A hybrid heuristic for the maximum clique problem, J. Heuristics 12 (1–2) (2006) 5–22.
    Google ScholarLocate open access versionFindings
  • W. Pullan, H.H. Hoos, Dynamic local search for the maximum clique problem, J. Artif. Intell. Res. (2006) 159–185.
    Google ScholarLocate open access versionFindings
  • W. Pullan, Phased local search for the maximum clique problem, J. Comb. Optim. 12 (3) (2006) 303–323.
    Google ScholarLocate open access versionFindings
  • P. Guturu, R. Dantu, An impatient evolutionary algorithm with probabilistic tabu search for unified solution of some NP-hard problems in graph and set theory via clique finding, IEEE Trans. Syst. Man Cybern., Part B, Cybern. 38 (3) (2008) 645–666.
    Google ScholarLocate open access versionFindings
  • Q. Wu, J.-K. Hao, An adaptive multistart tabu search approach to solve the maximum clique problem, J. Comb. Optim. 26 (1) (2013) 86–108.
    Google ScholarLocate open access versionFindings
  • U. Benlic, J.-K. Hao, Breakout local search for maximum clique problems, Comput. Oper. Res. 40 (1) (2013) 192–206.
    Google ScholarLocate open access versionFindings
  • I.M. Bomze, M. Pelillo, V. Stix, Approximating the maximum weight clique using replicator dynamics, IEEE Trans. Neural Netw. 11 (6) (2000) 1228–1241.
    Google ScholarLocate open access versionFindings
  • S. Busygin, A new trust region technique for the maximum weight clique problem, Discrete Appl. Math. 154 (15) (2006) 2080–2096.
    Google ScholarLocate open access versionFindings
  • W. Pullan, Approximating the maximum vertex/edge weighted clique using local search, J. Heuristics 14 (2) (2008) 117–134.
    Google ScholarLocate open access versionFindings
  • Q. Wu, J.-K. Hao, F. Glover, Multi-neighborhood tabu search for the maximum weight clique problem, Ann. Oper. Res. 196 (1) (2012) 611–634.
    Google ScholarLocate open access versionFindings
  • S. Cai, J. Lin, Fast solving maximum weight clique problem in massive graphs, in: Proceedings of the Twenty-Fifth International Joint Conference on Artificial Intelligence, IJCAI, 2016, pp. 568–574.
    Google ScholarLocate open access versionFindings
  • B. Nogueira, R.G. Pinheiro, A CPU-GPU local search heuristic for the maximum weight clique problem on massive graphs, Comput. Oper. Res. 90 (2018)
    Google ScholarLocate open access versionFindings
  • Y. Fan, N. Li, C. Li, Z. Ma, L.J. Latecki, K. Su, Restart and random walk in local search for maximum vertex weight cliques with evaluations in clustering aggregation, in: IJCAI, 2017, pp. 622–630.
    Google ScholarFindings
  • S. Cai, K. Su, A. Sattar, Local search with edge weighting and configuration checking heuristics for minimum vertex cover, Artif. Intell. 175 (9) (2011)
    Google ScholarLocate open access versionFindings
  • R. Li, S. Cai, S. Hu, M. Yin, J. Gao, NuMWVC: a novel local search for minimum weighted vertex cover problem, in: Proceedings of the Thirty-Second AAAI Conference on Artificial Intelligence, 2018, pp. 8107–8108.
    Google ScholarLocate open access versionFindings
  • Y. Wang, S. Cai, M. Yin, Local search for minimum weight dominating set with two-level configuration checking and frequency based scoring function, J. Artif. Intell. Res. 58 (2017) 267–295.
    Google ScholarLocate open access versionFindings
  • Y. Wang, S. Cai, J. Chen, M. Yin, A fast local search algorithm for minimum weight dominating set problem on massive graphs, in: IJCAI, 2018, pp. 1514–1522.
    Google ScholarFindings
  • Y. Wang, D. Ouyang, L. Zhang, M. Yin, A novel local search for unicost set covering problem using hyperedge configuration checking and weight diversity, Sci. China Inf. Sci. 60 (6) (2017) 062103.
    Google ScholarLocate open access versionFindings
  • Y. Wang, C. Li, H. Sun, J. Chen, M. Yin, MLQCC: an improved local search algorithm for the set k-covering problem, Int. Trans. Oper. Res. 26 (3) (2019)
    Google ScholarLocate open access versionFindings
  • S. Cai, K. Su, Configuration checking with aspiration in local search for sat, in: Proceedings of AAAI 2012, 2012, pp. 334–340.
    Google ScholarLocate open access versionFindings
  • S. Cai, K. Su, Local search for Boolean satisfiability with configuration checking and subscore, Artif. Intell. 204 (2013) 75–98.
    Google ScholarLocate open access versionFindings
  • A. Abramé, D. Habet, D. Toumi, Improving configuration checking for satisfiable random k-SAT instances, Ann. Math. Artif. Intell. 79 (1–3) (2017) 5–24.
    Google ScholarLocate open access versionFindings
  • C. Luo, S. Cai, W. Wu, Z. Jie, K. Su, CCLS: an efficient local search algorithm for weighted maximum satisfiability, IEEE Trans. Comput. 64 (7) (2015)
    Google ScholarLocate open access versionFindings
  • D.S. Johnson, Cliques, coloring, and satisfiability: second DIMACS implementation challenge, DIMACS Ser. Discret. Math. Theor. Comput. Sci. 26 (1993)
    Google ScholarLocate open access versionFindings
  • K. Xu, F. Boussemart, F. Hemery, C. Lecoutre, et al., A simple model to generate hard satisfiable instances, in: Proceedings of IJCAI, 2005, pp. 337–342.
    Google ScholarLocate open access versionFindings
  • Q. Wu, J.-K. Hao, A clique-based exact method for optimal winner determination in combinatorial auctions, Inf. Sci. 334 (2016) 103–121.
    Google ScholarLocate open access versionFindings
  • S. Cai, Balance between complexity and quality: local search for minimum vertex cover in massive graphs, in: Proceedings of IJCAI 2015, 2015, pp. 747–753.
    Google ScholarLocate open access versionFindings
  • R.A. Rossi, N.K. Ahmed, The network data repository with interactive graph analytics and visualization, in: Twenty-Ninth AAAI Conference on Artificial Intelligence, 2015, pp. 4292–4293.
    Google ScholarLocate open access versionFindings
  • Y. Wang, S. Cai, M. Yin, Two efficient local search algorithms for maximum weight clique problem, in: AAAI, 2016, pp. 805–811.
    Google ScholarFindings
  • C.M. Li, W. Huang, Diversification and determinism in local search for satisfiability, in: Proceedings of Theory and Applications of Satisfiability Testing, SAT 2005, 2005, pp. 158–172.
    Google ScholarLocate open access versionFindings
  • F. Glover, Tabu search – part I, ORSA J. Comput. 1 (3) (1989) 190–206.
    Google ScholarLocate open access versionFindings
  • Y. Zhou, J.-K. Hao, A. Goëffon Push, A generalized operator for the maximum vertex weight clique problem, Eur. J. Oper. Res. 257 (1) (2017) 41–54.
    Google ScholarLocate open access versionFindings
  • S. Cai, J. Lin, C. Luo, Finding a small vertex cover in massive sparse graphs: construct, local search, and preprocess, J. Artif. Intell. Res. 59 (2017)
    Google ScholarLocate open access versionFindings
  • J. Lin, S. Cai, C. Luo, K. Su, A reduction based method for coloring very large graphs, in: Twenty-Sixth International Joint Conference on Artificial Intelligence, 2017, pp. 517–523.
    Google ScholarLocate open access versionFindings
  • S. Cai, K. Su, C. Luo, A. Sattar, NuMVC: an efficient local search algorithm for minimum vertex cover, J. Artif. Intell. Res. (2013) 687–716.
    Google ScholarLocate open access versionFindings
  • Z. Ma, Y. Fan, K. Su, C. Li, A. Sattar, Deterministic tournament selection in local search for maximum edge weight clique on large sparse graphs, in: Australasian Joint Conference on Artificial Intelligence, Springer, 2017, pp. 353–364.
    Google ScholarLocate open access versionFindings
  • R.A. Rossi, D.F. Gleich, A.H. Gebremedhin, M.M.A. Patwary, Fast maximum clique algorithms for large graphs, in: Proceedings of the Companion Publication of the 23rd International Conference on World Wide Web Companion, 2014, pp. 365–366.
    Google ScholarLocate open access versionFindings
  • K. Xu, F. Boussemart, F. Hemery, C. Lecoutre, Random constraint satisfaction: easy generation of hard (satisfiable) instances, Artif. Intell. 171 (8) (2007)
    Google ScholarLocate open access versionFindings
  • K. Xu, W. Li, Exact phase transitions in random constraint satisfaction problems, J. Artif. Intell. Res. 12 (1) (2000) 93–103.
    Google ScholarLocate open access versionFindings
  • K. Xu, W. Li, Many hard examples in exact phase transitions, Theor. Comput. Sci. 355 (3) (2006) 291–302.
    Google ScholarLocate open access versionFindings
  • R. Stern, G. Dreiman, R. Valenzano, Probably bounded suboptimal heuristic search, Artif. Intell. 267 (2019) 39–57.
    Google ScholarLocate open access versionFindings
  • Z. Li, Q. Chen, V. Koltun, Combinatorial optimization with graph convolutional networks and guided tree search, in: Advances in Neural Information Processing Systems, 2018, pp. 537–546.
    Google ScholarLocate open access versionFindings
Your rating :
0

 

Tags
Comments