A Constrained Alternating Minimization Approach to End-to-End Graph Signal Denoising Over Unknown Topology

• Item type: SIG Technical Reports(1)
• 公開日: 2021-02-24
• タイトル: A Constrained Alternating Minimization Approach to End-to-End Graph Signal Denoising Over Unknown Topology
• タイトル / 言語: en
• タイトル: A Constrained Alternating Minimization Approach to End-to-End Graph Signal Denoising Over Unknown Topology
• 言語: eng
• キーワード / 主題Scheme: Other
• キーワード / 主題: SIP1
• 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: School of Computing, Tokyo Institute of Technology
• 著者所属: School of Computing, Tokyo Institute of Technology
• 著者所属(英): en
• 著者所属(英): School of Computing, Tokyo Institute of Technology
• 著者所属(英): en
• 著者所属(英): School of Computing, Tokyo Institute of Technology
• 著者名: Eisuke, Yamagata; Shunsuke, Ono
• 著者名(英): Eisuke, Yamagata; Shunsuke, Ono
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: This paper proposes a denoising method for smooth graph signals observed on a graph of unknown topology. The proposed method is motivated by GL-SigRep, which can be in interpreted as a denoising method considering that it outputs a denoised signal as well as the learned laplacian. Although GL-SigRep is an effective alternative optimization algorithm for graph learning, its denoising performance is not always optimal due to its regularization parameter, which appears in both denoising and graph learning phase of optimization, making parameter tuning difficult. In this paper, we modify GL-SigRep’s formulation to simplify its parameter tuning and illustrate the advantageous experimental results over synthetic data.
• 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 denoising method for smooth graph signals observed on a graph of unknown topology. The proposed method is motivated by GL-SigRep, which can be in interpreted as a denoising method considering that it outputs a denoised signal as well as the learned laplacian. Although GL-SigRep is an effective alternative optimization algorithm for graph learning, its denoising performance is not always optimal due to its regularization parameter, which appears in both denoising and graph learning phase of optimization, making parameter tuning difficult. In this paper, we modify GL-SigRep’s formulation to simplify its parameter tuning and illustrate the advantageous experimental results over synthetic data.
• 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: AN10442647
• 書誌情報: 研究報告音声言語情報処理（SLP）; 巻 2021-SLP-136, 号 3, p. 1-4, 発行日 2021-02-24
• ISSN / 収録物識別子タイプ: ISSN
• ISSN / 収録物識別子: 2188-8663
• 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