Hierarchical Importance Sampling as Generalized Population Convergence

• Item type: SIG Technical Reports(1)
• 公開日: 2007-09-03
• タイトル: Hierarchical Importance Sampling as Generalized Population Convergence
• タイトル / 言語: en
• タイトル: Hierarchical Importance Sampling as Generalized Population Convergence
• 言語: eng
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_18gh
• 資源タイプ: technical report
• 著者所属: Graduate School of Interdisciplinary Science and Engineering Tokyo Institute of Technology
• 著者所属: The University of Electro-Communications Faculty of Electro-communication
• 著者所属(英): en
• 著者所属(英): Graduate School of Interdisciplinary Science and Engineering, Tokyo Institute of Technology
• 著者所属(英): en
• 著者所属(英): The University of Electro-Communications, Faculty of Electro-communication
• 著者名: Takayuki, Higo; Keiki, Takadama
• 著者名(英): Takayuki, Higo; Keiki, Takadama
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: This paper proposes a novel method named Hierarchical Importance Sampling (HIS) as a generalization of the population convergence which plays an important role in Optimization Methods based on Probability Models (OMPM) such as Estimation of Distribution Algorithms and Cross Entropy methods. In HIS multiple populations are maintained simultaneously so that they have different diversities. HIS builds the probability models of the populations through importance sampling by using mix of the populations. Experimental comparisons between HIS and general OMPM have revealed that HIS outperforms general OMPM.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: This paper proposes a novel method, named Hierarchical Importance Sampling (HIS), as a generalization of the population convergence, which plays an important role in Optimization Methods based on Probability Models (OMPM) such as Estimation of Distribution Algorithms and Cross Entropy methods. In HIS, multiple populations are maintained simultaneously so that they have different diversities. HIS builds the probability models of the populations through importance sampling by using mix of the populations. Experimental comparisons between HIS and general OMPM have revealed that HIS outperforms general OMPM.
• 書誌レコードID / 収録物識別子タイプ: NCID
• 書誌レコードID / 収録物識別子: AN10505667
• 書誌情報: 情報処理学会研究報告数理モデル化と問題解決（MPS）; 巻 2007, 号 86(2007-MPS-066), p. 17-20, 発行日 2007-09-03
• Notice: SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc.
• 出版者 / 言語: ja
• 出版者: 情報処理学会