東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

FindBook

Google Book

Amazon

博客來

Learning and Signal Processing over High-Dimensional Graphs.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Learning and Signal Processing over High-Dimensional Graphs./
作者:	Zhang, Songyang.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2021,
面頁冊數:	194 p.
附註:	Source: Dissertations Abstracts International, Volume: 83-02, Section: B.
Contained By:	Dissertations Abstracts International83-02B.
標題:	Engineering. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28499457
ISBN:	9798535574639

Learning and Signal Processing over High-Dimensional Graphs.
Zhang, Songyang.

Learning and Signal Processing over High-Dimensional Graphs. - Ann Arbor : ProQuest Dissertations & Theses, 2021 - 194 p.

Source: Dissertations Abstracts International, Volume: 83-02, Section: B.

Thesis (Ph.D.)--University of California, Davis, 2021.

This item must not be sold to any third party vendors.

Learning and signal processing methods over graphs have recently attracted significant attentions in dealing with structured data. Normal (traditional) graphs, however, only capture pairwise relationships among nodes and are not effective in representing and capturing some high-order relationships of data samples. Such high order data interactions are often important in many applications such as Internet of Things (IoT), multimedia processing and network analysis, thereby motivating the exploration of learning and signal processing through high-dimensional graphs. In this dissertation, we investigate theoretical foundations and practical applications of two different high-dimensional graphs: 1) multilayer networks, and 2) hypergraphs. Inspired by the properties of high-dimensional graphs, we also revisit certain aspects of signal processing and learning under normal graphs.First, we study the behavior analysis of propagation over multilayer networks. Specifically, we focus on the cascading failure over multilayer complex systems. We first propose a scalable tensor-based framework to represent interdependent multilayer networks, before applying this framework to analyze the failure propagation based on a susceptible-infectious-susceptible (SIS) epidemic model. We derive the transition equations and failure threshold to characterize the failure propagation. To make the failure indicator analytically tractable and computationally efficient, we derive its upper and lower bounds, as well as its approximated expressions in special cases.Second, we investigate signal processing over hypergraphs. Representing hypergraphs as tensors, we proposed a novel framework of hypergraph signal processing (HGSP). Defining a specific form of hypergraph signals and hypergraph signal shifting, we provide an alternative definition of hypergraph Fourier space based on the tensor decomposition, together with the corresponding hypergraph Fourier transform. To better interpret the hypergraph Fourier space, we analyze the resulting hypergraph frequency properties, including the concepts of frequency and bandlimited signals. We also establish theoretical foundation for the HGSP sampling theory and filter designs. Furthermore, we examine its applications in multimedia processing, including three-dimensional (3D) point clouds, images and videos.Lastly, revisiting the traditional graphs, we investigate the development of graph convolutional networks from the perspective of graph signal processing (GSP). We reexamine the graph spectral convolution in GSP and define conditions for approximating spectrum wavelet via propagation in the vertex domain. We then propose alternative propagation models for the GCN layers and develop a Taylor-series based graph convolutional networks (TGCN) based on the aforementioned approximation conditions. Our experimental results in citation networks and point clouds validate the effectiveness of the proposed TGCN.

ISBN: 9798535574639Subjects--Topical Terms:

586835
Engineering.
Subjects--Index Terms:

Signal Processing

Learning and Signal Processing over High-Dimensional Graphs.
LDR:04045nmm a2200349 4500 001 2343208
005 20220502104201.5
008 241004s2021 ||||||||||||||||| ||eng d
020 $a 9798535574639
035 $a (MiAaPQ)AAI28499457
035 $a AAI28499457
040 $a MiAaPQ $c MiAaPQ
100 1 $a Zhang, Songyang. $3 3681695
245 1 0 $a Learning and Signal Processing over High-Dimensional Graphs.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2021
300 $a 194 p.
500 $a Source: Dissertations Abstracts International, Volume: 83-02, Section: B.
500 $a Advisor: Ding, Zhi.
502 $a Thesis (Ph.D.)--University of California, Davis, 2021.
506 $a This item must not be sold to any third party vendors.
520 $a Learning and signal processing methods over graphs have recently attracted significant attentions in dealing with structured data. Normal (traditional) graphs, however, only capture pairwise relationships among nodes and are not effective in representing and capturing some high-order relationships of data samples. Such high order data interactions are often important in many applications such as Internet of Things (IoT), multimedia processing and network analysis, thereby motivating the exploration of learning and signal processing through high-dimensional graphs. In this dissertation, we investigate theoretical foundations and practical applications of two different high-dimensional graphs: 1) multilayer networks, and 2) hypergraphs. Inspired by the properties of high-dimensional graphs, we also revisit certain aspects of signal processing and learning under normal graphs.First, we study the behavior analysis of propagation over multilayer networks. Specifically, we focus on the cascading failure over multilayer complex systems. We first propose a scalable tensor-based framework to represent interdependent multilayer networks, before applying this framework to analyze the failure propagation based on a susceptible-infectious-susceptible (SIS) epidemic model. We derive the transition equations and failure threshold to characterize the failure propagation. To make the failure indicator analytically tractable and computationally efficient, we derive its upper and lower bounds, as well as its approximated expressions in special cases.Second, we investigate signal processing over hypergraphs. Representing hypergraphs as tensors, we proposed a novel framework of hypergraph signal processing (HGSP). Defining a specific form of hypergraph signals and hypergraph signal shifting, we provide an alternative definition of hypergraph Fourier space based on the tensor decomposition, together with the corresponding hypergraph Fourier transform. To better interpret the hypergraph Fourier space, we analyze the resulting hypergraph frequency properties, including the concepts of frequency and bandlimited signals. We also establish theoretical foundation for the HGSP sampling theory and filter designs. Furthermore, we examine its applications in multimedia processing, including three-dimensional (3D) point clouds, images and videos.Lastly, revisiting the traditional graphs, we investigate the development of graph convolutional networks from the perspective of graph signal processing (GSP). We reexamine the graph spectral convolution in GSP and define conditions for approximating spectrum wavelet via propagation in the vertex domain. We then propose alternative propagation models for the GCN layers and develop a Taylor-series based graph convolutional networks (TGCN) based on the aforementioned approximation conditions. Our experimental results in citation networks and point clouds validate the effectiveness of the proposed TGCN.
590 $a School code: 0029.
650 4 $a Engineering. $3 586835
650 4 $a Artificial intelligence. $3 516317
650 4 $a Eigen values. $3 3680699
650 4 $a Propagation. $3 3680519
650 4 $a Approximation. $3 3560410
650 4 $a Accuracy. $3 3559958
650 4 $a Datasets. $3 3541416
650 4 $a Multimedia. $2 gtt $3 1000510
650 4 $a Data compression. $3 3681696
650 4 $a Graph representations. $3 3560730
650 4 $a Signal processing. $3 533904
653 $a Signal Processing
653 $a Graphs
653 $a Hypergraphs
653 $a Machine learning
690 $a 0537
690 $a 0800
710 2 $a University of California, Davis. $b Electrical and Computer Engineering. $3 1672487
773 0 $t Dissertations Abstracts International $g 83-02B.
790 $a 0029
791 $a Ph.D.
792 $a 2021
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28499457