Cross Entropy Clustering Algorithm Based on Transfer Learning
To solve the problem of clustering performance degradation when traditional clustering algorithms are applied to insufficient or noisy data, a cross entropy clustering algorithm based on transfer learning is proposed. It improves the classical cross entropy clustering algorithm by combining knowledges from historical clustering centers and historical degree of membership and applying them to the objective function proposed for clustering insufficient or noisy target data. The experiment results on several synthetic and four real datasets and analyses show the proposed algorithm has high effectiveness over the available.
KeywordsCluster Cross entropy clustering Transfer learning Historical clustering center Historical degree of membership
This work was supported in part by the National Natural Science Foundation of China under Grants (61472307, 51405387), the Key Research Project of Shaanxi Province (2018GY-018) and the Foundation of Education Department of Shaanxi Province (17JK0713).
- 2.Wang, L., Wang, J.D., Li, T.: Cluster’s feature weighting fuzzy clustering algorithm integrating rough sets and shadowed sets. Syst. Eng. Electron. 35(8), 1769–1776 (2013). https://doi.org/10.3969/j.issn.1001-506X.2013.08.31CrossRefzbMATHGoogle Scholar
- 6.Li, C.M., Xu, S.B., Hao, Z.F.: Cross-entropy semi-supervised clustering based on pairwise on constraints. CAAI Trans. Intell. Syst. 30(07), 598–608 (2017). https://doi.org/10.16451/j.cnki.issn1003-6059.201707003CrossRefGoogle Scholar
- 8.Long, M., Zhou, T.J.: A survey of properties and application of the Lambert W function. J. Hengyang Norm. Univ. 32(6), 38–40 (2011). https://doi.org/10.13914/j.cnki.cn43-1453/z.2011.06.010MathSciNetCrossRefGoogle Scholar