HSICを拡張した条件付き独立検定

• Item type: SIG Technical Reports(1)
• 公開日: 2022-06-20
• タイトル: HSICを拡張した条件付き独立検定
• タイトル / 言語: en
• タイトル: Extending HSIC for Testing Conditional Independence
• 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: 大阪大学基礎工学研究科
• 著者所属: 大阪大学基礎工学研究科
• 著者所属(英): en
• 著者所属(英): Graduate School of Engineering Science, Osaka University
• 著者所属(英): en
• 著者所属(英): Graduate School of Engineering Science, Osaka University
• 著者名: 張, 秉元; 鈴木, 讓
• 著者名(英): Bingyuan, Zhang; Joe, Suzuki
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: Conditional Independence (CI) testing is a fundamental problem in statistics, which is applied directly to causal discovery. Many nonparametric CI tests are developed, but a common challenge exists: current methods perform poorly with a high dimensional conditioning set. To tackle this problem, we consider a novel nonparametric CI test using a kernel-based measure, which can be viewed as an extension of the Hilbert-Schmidt Independence Criterion (HSIC). The experimental results show that our proposed method leads to a significant performance improvement when compared with previous methods. In particular, our method performs well against the growth of the dimension of the conditioning set. Meanwhile, our method shows competitive scalability regarding the sample size ???? and the dimension of the conditioning set.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: Conditional Independence (CI) testing is a fundamental problem in statistics, which is applied directly to causal discovery. Many nonparametric CI tests are developed, but a common challenge exists: current methods perform poorly with a high dimensional conditioning set. To tackle this problem, we consider a novel nonparametric CI test using a kernel-based measure, which can be viewed as an extension of the Hilbert-Schmidt Independence Criterion (HSIC). The experimental results show that our proposed method leads to a significant performance improvement when compared with previous methods. In particular, our method performs well against the growth of the dimension of the conditioning set. Meanwhile, our method shows competitive scalability regarding the sample size n and the dimension of the conditioning set.
• 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: 研究報告数理モデル化と問題解決（MPS）; 巻 2022-MPS-138, 号 29, p. 1-6, 発行日 2022-06-20
• ISSN / 収録物識別子タイプ: ISSN
• ISSN / 収録物識別子: 2188-8833
• 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