DOCSLIB.ORG

Fim019-Endmatter.Pdf

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Fim019-Endmatter.Pdf Efficient Grap h Representations This page intentionally left blank http://dx.doi.org/10.1090/fim/019 FIELDS INSTITUT E MONOGRAPHS THE FIELD S INSTITUT E FO R RESEARC H I N MATHEMATICA L SCIENCE S Efficient Grap h Representations Jeremy P . Spinra d American Mathematica l Societ y Providence, Rhod e Islan d The Field s Institut e for Researc h i n Mathematical Science s The Field s Institute i s named i n honour o f the Canadia n mathematicia n Joh n Charle s Fields (1863-1932) . Field s wa s a visionar y wh o receive d man y honour s fo r hi s scientifi c work, including election to the Royal Society o f Canada i n 190 9 and to the Royal Society of London i n 1913 . Amon g othe r accomplishment s i n the servic e o f the internationa l math - ematics community , Field s wa s responsibl e fo r establishin g th e world' s mos t prestigiou s prize fo r mathematic s research—th e Field s Medal . The Fields Institute fo r Researc h i n Mathematical Science s i s supported b y grants fro m the Ontario Ministry o f Education an d Trainin g and the Natural Science s and Engineerin g Research Counci l o f Canada . Th e Institut e i s sponsore d b y McMaste r University , th e University o f Toronto, the Universit y o f Waterloo, an d Yor k University , an d ha s affiliate d universities i n Ontari o an d acros s Canada . 2000 Mathematics Subject Classification. Primar y 05C62 , 05C17 , 05C50 , 05C85 , 05-00, 05-02, 68R10 , 68W01 , 68P05 , 68Q30, 68-01 . For additiona l informatio n an d update s o n this book , visi t www.ams.org/bookpages/fim-19 Library o f Congress Cataloging-in-Publicatio n Dat a Spinrad, Jerem y P. Efficient grap h representation s / Jerem y P . Spinrad . p. cm. — (Fields Institut e monograph s ; 19) Includes bibliographica l reference s an d index. ISBN 0-8218-2815- 0 (acid-fre e paper ) 1. Representations o f graphs. I . Title. II . Series. QA166.242.S65 200 3 200304510 6 511/.-dc21 CI P Copying an d reprinting . Individua l reader s o f this publication , an d nonprofi t librarie s acting fo r them, ar e permitted t o make fai r us e of the material, suc h as to copy a chapter fo r use in teachin g o r research . Permissio n i s granted t o quot e brie f passage s fro m thi s publicatio n i n reviews, provide d the customary acknowledgmen t o f the source i s given. Republication, systemati c copying , or multiple reproduction o f any material in this publicatio n is permitte d onl y unde r licens e fro m th e America n Mathematica l Society . Request s fo r suc h permission shoul d be addressed to the Acquisitions Department , America n Mathematica l Society , 201 Charle s Street , Providence , Rhod e Islan d 02904-2294 , USA . Requests ca n also b e made by e-mail to [email protected] . © 200 3 by the American Mathematica l Society . Al l rights reserved . The America n Mathematica l Societ y retain s al l rights except thos e grante d to the United State s Government . Printed i n the United State s o f America. @ Th e paper use d i n this boo k i s acid-free an d falls withi n the guidelines established t o ensure permanenc e an d durability. This publicatio n wa s prepared b y The Fields Institute . Visit the AMS home pag e at http: //www. ams. org/ 10 9 8 7 6 5 4 3 2 1 0 8 07 06 05 04 03 Contents Explanatory Remark s 1 Chapter 1 . Introductio n 5 1.1. Grap h Theor y Backgroun d an d Terminolog y 6 1.2. Algorith m Backgroun d an d Terminolog y 6 1.3. Representatio n Backgroun d 8 1.4. Exampl e o f a Nic e Representation 1 1 1.5. Overvie w o f Problems i n Graph Representatio n 1 2 1.6. Exercise s 1 5 Chapter 2 . Implici t Representatio n 1 7 2.1. Implici t Representatio n an d Universa l Graph s 2 1 2.2. Generalize d Implici t Representatio n 2 2 2.3. Representation s wit h Ver y Shor t Label s 2 3 2.4. Distanc e Labelin g o f Graphs 2 5 2.5. Exercise s 2 7 Chapter 3 . Intersectio n an d Containmen t Representation s 3 1 3.1. Chordalit y an d Transitiv e Orientatio n 3 1 3.2. Technique s fo r Interva l Graph s 3 3 3.3. Generalization s o f Interval Graph s 3 4 3.4. Permutatio n Graph s an d Generalization s 3 8 3.5. Containmen t Representation s 4 1 3.6. Overla p Representation s 4 2 3.7. Generalize d Intersectio n Model s 4 4 3.8. Perfec t Graph s 4 6 3.9. Exercise s 4 7 Chapter 4 . Rea l Numbers i n Graph Representation s 5 3 4.1. Warren' s Theore m 5 4 4.2. Continuou s Nongeometri c Variable s 5 6 4.3. Decidabilit y Result s 5 7 4.4. Exercise s 5 7 Chapter 5 . Classe s Which us e Globa l Informatio n 5 9 5.1. Path s i n Trees 5 9 5.2. Chorda l Comparabilit y Graph s 6 0 5.3. Fill-i n Scheme s 6 5 5.4. Closur e Operation s 6 6 5.5. Weakl y Chorda l Graph s 6 7 vi Content s 5.6. Othe r Fill-i n Scheme s 6 8 5.7. Exercise s 6 8 Chapter 6 . Visibilit y Graph s 7 3 6.1. Countin g Visibilit y Graph s 7 4 6.2. Cliqu e Cove r Representatio n 7 4 6.3. Induce d Visibilit y Graph s 7 6 6.4. Optimizatio n o n Visibility Graph s 8 0 6.5. Lin e o f Sigh t Graph s an d Othe r Variant s 8 0 6.6. Exercise s 8 1 Chapter 7 . Intersectio n o f Graph Classe s 8 5 7.1. Som e Fundamental Propertie s 8 5 7.2. Intersection s o f Fundamental Propertie s 8 7 7.3. Weakl y Chorda l Comparabilit y Graph s 8 8 7.4. Othe r Generalize d Classe s 9 0 7.5. Observation s Regardin g AT-fre e co-AT-fre e Graph s 9 1 7.6. Ope n Problem s o n the Generalize d Classe s 9 4 7.7. Exercise s 9 5 Chapter 8 . Grap h Classe s Define d b y Forbidden Subgraph s 9 7 8.1. Cograph s 9 7 8.2. Classe s Which ar e too Larg e to hav e Efficien t Representation s 10 0 8.3. Relatio n Betwee n Recognitio n Problem s 10 5 8.4. Classe s define d b y Forbidding Set s o f Induced Subgraph s 10 5 8.5. Dilwort h Numbe r an d Pose t Widt h 10 7 8.6. Exercise s 10 9 Chapter 9 . Chorda l Bipartit e Graph s 11 1 9.1. T-fre e Matrice s 11 1 9.2. Countin g an d Representatio n 11 2 9.3. Characterization s 11 8 9.4. Recognitio n 12 0 9.5. Optimizatio n Problem s o n Chorda l Bipartit e Graph s 12 3 9.6. Variant s o f Chordal Bipartit e Graph s 12 5 9.7. Subclasse s o f Chordal Bipartit e Graph s 12 6 9.8. Perfec t Eliminatio n Bipartit e Graph s 13 0 9.9. Bipartit e Graph s wit h Forbidde n Induce d Subgraph s 13 1 9.10. Exercise s 13 2 Chapter 10 . Matrice s 13 5 10.1. Smal l Forbidden Classe s o f Matrices 13 5 10.2. Linea r Matrice s 13 6 10.3. Forbidde n 2 by 2 Identity Matrice s 13 8 10.4. Forbiddin g (™ ) 13 9 10.5. Othe r Classe s o f Interest 14 2 10.6. Th e Problem s o f Counting an d Representatio n 14 2 10.7. Othe r Matri x Propertie s 14 5 10.8. Exercise s 14 5 Chapter 11 . Decompositio n 149 Contents vi i 11.1. Substitutio n Decompositio n an d Verte x Partitioning 14 9 11.2. Joi n Decompositio n 16 0 11.3. Recursivel y Decomposabl e Graph s 16 3 11.4. Clique-widt h an d NLC-widt h 16 4 11.5. Cliqu e Separato r Decompositio n 16 7 11.6. Ske w Partition 17 0 11.7. 2-Joi n 17 4 11.8. Exercise s 17 5 Chapter 12 . Eliminatio n Scheme s 18 1 12.1. Distanc e Hereditar y Graph s 18 1 12.2. Strongl y Chorda l Graph s 18 2 12.3. k-Simplicia l Eliminatio n Scheme s 18 4 12.4.
Load more

Recommended publications
  • A Complexity Theory for Implicit Graph Representations
    A Complexity Theory for Implicit Graph Representations Maurice Chandoo Leibniz Universität Hannover, Theoretical Computer Science, Appelstr. 4, 30167 Hannover, Germany [email protected] Abstract In an implicit graph representation the vertices of a graph are assigned short labels such that adjacency of two vertices can be algorithmically determined from their labels. This allows for a space-efficient representation of graph classes that have asymptotically fewer graphs than the class of all graphs such as forests or planar graphs. A fundamental question is whether every smaller graph class has such a representation. We consider how restricting the computational complexity of such representations affects what graph classes can be represented. A formal language can be seen as implicit graph representation and therefore complexity classes such as P or NP can be understood as set of graph classes. We investigate this complexity landscape and introduce complexity classes defined in terms of first order logic that capture surprisingly many of the graph classes for which an implicit representation is known. Additionally, we provide a notion of reduction between graph classes which reveals that trees and interval graphs are ‘complete’ for certain fragments of first order logic in this setting. 1998 ACM Subject Classification G.2.2 Graph Theory, F.1.3 Complexity Measures and Classes Keywords and phrases adjacency labeling schemes, complexity of graph properties We call a graph class small if it has at most 2O(n log n) graphs on n vertices. Many important graph classes such as interval graphs are small. Using adjacency matrices or lists to represent such classes is not space-efficient since this requires n2 bits (resp.
    [Show full text]
  • Estimating All Pairs Shortest Paths in Restricted Graph Families: a Unified Approach
    Estimating All Pairs Shortest Paths in Restricted Graph Families: A Unified Approach (Extended Abstract) Feodor F. Dragan Dept. of Computer Science, Kent State University, Kent, Ohio 44242, USA [email protected] Abstract. In this paper we show that a very simple and efficient ap- proach can be used to solve the all pairs almost shortest path problem on the class of weakly chordal graphs and its different subclasses. Moreover, this approach works well also on graphs with small size of largest induced cycle and gives a unified way to solve the all pairs almost shortest path and all pairs shortest path problems on different graph classes includ- ing chordal, strongly chordal, chordal bipartite, and distance-hereditary graphs. 1 Introduction Let G =(V,E) be a finite, unweighted, undirected, connected and simple (i.e., without loops and multiple edges) graph. Let also |V | = n and |E| = m. The distance d(v, u) between vertices v and u is the smallest number of edges in a path connecting v and u. The all pairs shortest path problem is to compute d(v, u) for all pairs of vertices v and u in G. The all pairs shortest path (APSP) problem is one of the most fundamental graph problems. There has been a renewed interest in it recently (see [1,2,4,6,12], [15,29] for general (unweighted, undirected) graphs, and [3,5,7,10,11,14,20,21,24], [28,32,33] for special graph classes). For general graphs, the best result known is by Seidel [29], who showed that the APSP problem can be solved in O(M(n) log n) time where M(n) denotes the time complexity for matrix multiplication involving small integers only.
    [Show full text]
  • Abstracting Abstraction in Search II: Complexity Analysis
    Proceedings of the Fifth Annual Symposium on Combinatorial Search Abstracting Abstraction in Search II: Complexity Analysis Christer Backstr¨ om¨ and Peter Jonsson Department of Computer Science, Linkoping¨ University SE-581 83 Linkoping,¨ Sweden [email protected] [email protected] Abstract In our earlier publication (Backstr¨ om¨ and Jonsson 2012) we have demonstrated the usefulness of this framework in Modelling abstraction as a function from the original state various ways. We have shown that a certain combination of space to an abstract state space is a common approach in com- our transformation properties exactly captures the DPP con- binatorial search. Sometimes this is too restricted, though, cept by Zilles and Holte (2010). A related concept is that and we have previously proposed a framework using a more flexible concept of transformations between labelled graphs. of path/plan refinement without backtracking to the abstract We also proposed a number of properties to describe and clas- level. Partial solutions to capturing this concept have been sify such transformations. This framework enabled the mod- presented in the literature, for instance, the ordered mono- elling of a number of different abstraction methods in a way tonicity criterion by Knoblock, Tenenberg, and Yang (1991), that facilitated comparative analyses. It is of particular inter- the downward refinement property (DRP) by Bacchus and est that these properties can be used to capture the concept of Yang (1994), and the simulation-based approach by Bundy refinement without backtracking between levels; how to do et al. (1996). However, how to define general conditions this has been an open question for at least twenty years.
    [Show full text]
  • Implicit Structures for Graph Neural Networks
    Implicit Structures for Graph Neural Networks Fangda Gu Electrical Engineering and Computer Sciences University of California at Berkeley Technical Report No. UCB/EECS-2020-185 http://www2.eecs.berkeley.edu/Pubs/TechRpts/2020/EECS-2020-185.html November 23, 2020 Copyright © 2020, by the author(s). All rights reserved. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission. Acknowledgement I would like to express my most sincere gratitude to Professor Laurent El Ghaoui and Professor Somayeh Sojoudi for their support in my academic life. The helpful discussions with them on my research have guided me through the past two successful years at University of California, Berkeley. I want to thank Heng Chang and Professor Wenwu Zhu for their help in the implementation and verification of the work. I also want to thank Shirley Salanio for her strong logistical and psychological support. Implicit Structures for Graph Neural Networks by Fangda Gu Research Project Submitted to the Department of Electrical Engineering and Computer Sciences, University of California at Berkeley, in partial satisfaction of the requirements for the degree of Master of Science, Plan II. Approval for the Report and Comprehensive Examination: Committee: Professor Laurent El Ghaoui Research Advisor (Date) * * * * * * * Professor Somayeh Sojoudi Second Reader 11/19/2020 (Date) Implicit Structures for Graph Neural Networks Fangda Gu Abstract Graph Neural Networks (GNNs) are widely used deep learning models that learn meaningful representations from graph-structured data.
    [Show full text]
  • Minimum Fill-In for Chordal Bipartite Graphs
    Minimum fill-in for chordal bipartite graphs Citation for published version (APA): Kloks, A. J. J. (1993). Minimum fill-in for chordal bipartite graphs. (Universiteit Utrecht. UU-CS, Department of Computer Science; Vol. 9311). Utrecht University. Document status and date: Published: 01/01/1993 Document Version: Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers) Please check the document version of this publication: • A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website. • The final author version and the galley proof are versions of the publication after peer review. • The final published version features the final layout of the paper including the volume, issue and page numbers. Link to publication General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal.
    [Show full text]
  • Large-Scale Directed Model Checking LTL
    Large-Scale Directed Model Checking LTL Stefan Edelkamp and Shahid Jabbar University of Dortmund Otto-Hahn Straße 14 {stefan.edelkamp,shahid.jabbar}@cs.uni-dortmund.de Abstract. To analyze larger models for explicit-state model checking, directed model checking applies error-guided search, external model check- ing uses secondary storage media, and distributed model checking exploits parallel exploration on multiple processors. In this paper we propose an external, distributed and directed on-the-fly model checking algorithm to check general LTL properties in the model checker SPIN. Previous attempts restricted to checking safety proper- ties. The worst-case I/O complexity is bounded by O(sort(|F||R|)/p + l · scan(|F||S|)), where S and R are the sets of visited states and transitions in the synchronized product of the B¨uchi automata for the model and the property specification, F is the number of accepting states, l is the length of the shortest counterexample, and p is the number of processors. The algorithm we propose returns minimal lasso-shaped counterexam- ples and includes refinements for property-driven exploration. 1 Introduction The core limitation to the exploration of systems are bounded main memory resources. Relying on virtual memory slows down the exploration due to excessive page faults. External algorithms [31] exploit hard disk space and organize the access to secondary memory. Originally designed for explicit graphs, external search algorithms have shown considerably good performances in the large-scale breadth-first and guided exploration of games [22, 12] and in the analysis of model checking problems [24]. Directed explicit-state model checking [13] enhances the error-reporting capa- bilities of model checkers.
    [Show full text]
  • Graph Isomorphism Completeness for Chordal Bipartite Graphs and Strongly Chordal Graphs
    Graph Isomorphism Completeness for Chordal bipartite graphs and Strongly Chordal Graphs Ryuhei Uehara a Seinosuke Toda b Takayuki Nagoya c aNatural Science Faculty, Komazawa University. 1 bDepartment of Computer Science and System Analysis, College of Humanities and Sciences, Nihon University. cDepartment of Mathematical Sciences, Tokyo Denki University. Abstract This paper deals with the graph isomorphism (GI) problem for two graph classes: chordal bipartite graphs and strongly chordal graphs. It is known that GI problem is GI complete even for some special graph classes including regular graphs, bipar- tite graphs, chordal graphs, comparability graphs, split graphs, and k-trees with unbounded k. On the other side, the relative complexity of the GI problem for the above classes was unknown. We prove that deciding isomorphism of the classes are GI complete. Key words: Graph isomorphism problem, Graph isomorphism complete, Strongly chordal graphs, Chordal bipartite graphs 1 Introduction The graph isomorphism (GI) problem is a well-known problem and exploring its precise complexity has become an important open question in computa- tional complexity theory three decades ago (see (Kar72)). Although the prob- lem is trivially in NP, the problem is not known to be in P and not known to be NP-complete either (see (RC77; KST93)). It is very unlikely that the Email addresses: [email protected] (Ryuhei Uehara), [email protected] (Seinosuke Toda), [email protected] (Takayuki Nagoya). 1 This work was done while the author was visiting University of Waterloo. Preprint submitted to Elsevier Science 26 May 2004 GI problem is NP-complete.
    [Show full text]
  • NP-Hard Graph Problems and Boundary Classes of Graphs
    View metadata, citation and similar papers at core.ac.uk brought to you by CORE provided by Elsevier - Publisher Connector Theoretical Computer Science 389 (2007) 219–236 www.elsevier.com/locate/tcs NP-hard graph problems and boundary classes of graphs V.E. Alekseeva, R. Boliacb, D.V. Korobitsyna, V.V. Lozinb,∗ a University of Nizhny Novgorod, Gagarina 23, Nizhny Novgorod, 603950, Russia b RUTCOR, Rutgers University, 640 Bartholomew Road, Piscataway NJ, 08854-8003, USA Received 1 September 2004; received in revised form 7 February 2007; accepted 10 September 2007 Communicated by A. Razborov Abstract Any graph problem, which is NP-hard in general graphs, becomes polynomial-time solvable when restricted to graphs in special classes. When does a difficult problem become easy? To answer this question we study the notion of boundary classes. In the present paper we define this notion in its most general form, describe several approaches to identify boundary classes and apply them to various graph problems. c 2007 Elsevier B.V. All rights reserved. Keywords: Computational complexity; Hereditary class of graphs 1. Introduction We study graph problems that are NP-hard in general, i.e. in the class of all graphs. When we restrict the class of input graphs, any of these problems can become tractable, i.e. solvable in polynomial time. When does a difficult problem become easy? Is there any boundary that separates classes of graphs where the problem remains NP-hard from those where it has a polynomial-time solution? To answer these questions, we study the notion of boundary classes of graphs.
    [Show full text]
  • Cliques and Clubs⋆
    Cliques and Clubs? Petr A. Golovach1, Pinar Heggernes1, Dieter Kratsch2, and Arash Rafiey1 1 Department of Informatics, University of Bergen, Norway, fpetr.golovach,pinar.heggernes,[email protected] 2 LITA, Universit´ede Lorraine - Metz, France, [email protected] Abstract. Clubs are generalizations of cliques. For a positive integer s, an s-club in a graph G is a set of vertices that induces a subgraph of G of diameter at most s. The importance and fame of cliques are evident, whereas clubs provide more realistic models for practical applications. Computing an s-club of maximum cardinality is an NP-hard problem for every fixed s ≥ 1, and this problem has attracted significant attention recently. We present new positive results for the problem on large and important graph classes. In particular we show that for input G and s, a maximum s-club in G can be computed in polynomial time when G is a chordal bipartite or a strongly chordal or a distance hereditary graph. On a superclass of these graphs, weakly chordal graphs, we obtain a polynomial-time algorithm when s is an odd integer, which is best possible as the problem is NP-hard on this clas for even values of s. We complement these results by proving the NP-hardness of the problem for every fixed s on 4-chordal graphs, a superclass of weakly chordal graphs. Finally, if G is an AT-free graph, we prove that the problem can be solved in polynomial time when s ≥ 2, which gives an interesting contrast to the fact that the problem is NP-hard for s = 1 on this graph class.
    [Show full text]
  • Shorter Implicit Representation for Planar Graphs and Bounded Treewidth Graphs
    Shorter Implicit Representation for Planar Graphs and Bounded Treewidth Graphs Cyril Gavoille and Arnaud Labourel LaBRI, Universit´e de Bordeaux, France {gavoille,labourel}@labri.fr Abstract. Implicit representation of graphs is a coding of the structure of graphs using distinct labels so that adjacency between any two vertices can be decided by inspecting their labels alone. All previous implicit rep- resentations of planar graphs were based on the classical three forests de- composition technique (a.k.a. Schnyder’s trees), yielding asymptotically toa3logn-bit1 label representation where n is the number of vertices of the graph. We propose a new implicit representation of planar graphs using asymptotically 2 log n-bit labels. As a byproduct we have an explicit construction of a graph with n2+o(1) vertices containing all n-vertex pla- nar graphs as induced subgraph, the best previous size of such induced- universal graph was O(n3). More generally, for graphs excluding a fixed minor, we construct a 2logn + O(log log n) implicit representation. For treewidth-k graphs we give a log n + O(k log log(n/k)) implicit representation, improving the O(k log n) representation of Kannan, Naor and Rudich [18] (STOC ’88). Our representations for planar and treewidth-k graphs are easy to implement, all the labels can be constructed in O(n log n)time,and support constant time adjacency testing. 1 Introduction How to represent a graph in memory is a basic and fundamental data structure question. The two basic ways are adjacency matrices and adjacency lists. The latter representation is space efficient for sparse graphs, but adjacency queries require searching in the list, whereas matrices allow fast queries to the price of a super-linear space.
    [Show full text]
  • New Min-Max Theorems for Weakly Chordal and Dually Chordal Graphs
    New Min-Max Theorems for Weakly Chordal and Dually Chordal Graphs Arthur H. Busch1,, Feodor F. Dragan2, and R. Sritharan3, 1 Department of Mathematics, The University of Dayton, Dayton, OH 45469 [email protected] 2 Department of Computer Science, Kent State University, Kent, OH 44242 [email protected] 3 Department of Computer Science, The University of Dayton, Dayton, OH 45469 [email protected] Abstract. Adistance-k matching in a graph G is matching M in which the distance between any two edges of M is at least k.Adistance-2 matching is more commonly referred to as an induced matching. In this paper, we show that when G is weakly chordal, the size of the largest induced matching in G is equal to the minimum number of co-chordal subgraphs of G needed to cover the edges of G, and that the co-chordal subgraphs of a minimum cover can be found in polynomial time. Us- ing similar techniques, we show that the distance-k matching problem for k>1 is tractable for weakly chordal graphs when k is even, and is NP-hard when k is odd. For dually chordal graphs, we use properties of hypergraphs to show that the distance-k matching problem is solvable in polynomial time whenever k is odd, and NP-hard when k is even. Moti- vated by our use of hypergraphs, we define a class of hypergraphs which lies strictly in between the well studied classes of acyclic hypergraphs and normal hypergraphs. 1 Background and Motivation In this paper, all graphs are undirected, simple, and finite.
    [Show full text]
  • Graph Fairing Convolutional Networks for Anomaly Detection
    Graph Fairing Convolutional Networks for Anomaly Detection Mahsa Mesgaran and A. Ben Hamza Concordia Institute for Information Systems Engineering Concordia University, Montreal, QC, Canada Abstract class (i.e. normal class) by learning a discriminative hyperplane boundary around the normal instances. Another commonly- Graph convolution is a fundamental building block for many used anomaly detection approach is support vector data descrip- deep neural networks on graph-structured data. In this paper, tion (SVDD) [4], which basically finds the smallest possible hy- we introduce a simple, yet very effective graph convolutional persphere that contains all instances, allowing some instances to network with skip connections for semi-supervised anomaly de- be excluded as anomalies. However, these approaches rely on tection. The proposed multi-layer network architecture is theo- hand-crafted features, are unable to appropriately handle high- retically motivated by the concept of implicit fairing in geome- dimensional data, and often suffer from computational scalabil- try processing, and comprises a graph convolution module for ity issues. aggregating information from immediate node neighbors and Deep learning has recently emerged as a very powerful way a skip connection module for combining layer-wise neighbor- to hierarchically learn abstract patterns from data, and has been hood representations. In addition to capturing information from successfully applied to anomaly detection, showing promising distant graph nodes through skip connections between the net- results in comparison with shallow methods [5]. Ruff et al. [6] work’s layers, our approach exploits both the graph structure and extend the shallow one-class classification SVDD approach to node features for learning discriminative node representations.
    [Show full text]