DOCSLIB.ORG

Distance Metrics in Variant Graphs

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Distance Metrics in Variant Graphs Distance Metrics in Variant Graphs Martin Gjesdal Bjørndal Master’s Thesis, Spring 2017 This master’s thesis is submitted under the master’s programme Modelling and Data Analysis, with programme option Statistics and Data Analysis, at the Department of Mathematics, University of Oslo. The scope of the thesis is 60 credits. The front page depicts a section of the root system of the exceptional Lie group E8, projected into the plane. Lie groups were invented by the Norwegian mathematician Sophus Lie (1842–1899) to express symmetries in differential equations and today they play a central role in various parts of mathematics. Distance Metrics in Variant Graphs Supervisors: Author: Geir Olve Storvik Martin Gjesdal Bjørndal Lex Nederbragt Abstract The traditional, linear representation of the genetic information in pop- ulations cause a loss of data, as it can not fully represent sites where the sequences differ. Conversely, a graph may incorporate the variation, thereby providing a mean of utilizing more information in the analysis of genetic dif- ferences between populations. This thesis deals with the quantification of the differences between two population graphs. Three ways of measuring the distance between population graphs are presented. The first way counts unique variants and compares them to the total number of variants. The second calculates the graph edit distance. The third specifies two probability models regarding the genotype distribution at each variant, and then calculates the Bayes factor at each location. The three distance measures are tested on data describing variations in the human genome. Six populations of humans from distinct geographical areas are represented in the data. The distance measures seem to give similar conclusions about the relative distances among pairs of populations. In order to place the distances in a context, they are evaluated using permutation tests. The permutation tests report significant results for all pairs of populations except one. Additionally, this is the pair that is given the shortest distance by all distance measures. In general, the methods and ideas presented in this thesis allow more ge- netic information to be included in the study of relationships between groups. As such, it may prosper to become more accurate than traditional designs. Keywords: Bayes factor, graph edit distance, permutation test, variant graph. iii Preface The idea for this thesis originated from the observation that linear representa- tions of DNA from more than one individual is associated with a loss of infor- mation. A genetic variation among the individuals of a population means that there is more than one alternative present in the genomic sequence. A linear representation will in this case select one of the alternatives instead of using all the data available. This idea paved the road to graphs, which provides an intuitive extension of the work done with strings in previous literature. A demanding challenge has been the lack of previous work to support my own writings. The use of graphs for assembling DNA sequences is extensive. However, the representations of multiple sequences is still largely dependent on strings. Despite challenges, the thesis did at no point cease to be inter- esting. Rather the contrary. As time has progressed and insight deepened, I find myself profoundly interested in a problem that I hope to witness the applications of. Though my contribution to the field is modest, I hope that the ideas presented in this thesis may inspire others to continue the work on population graphs and how they relate to each other. I believe it to be a topic of great potential. The rationale behind this claim is simple: Why not use all the data? I wish to express my sincere gratitude to my supervisor Geir Storvik for his continuous support and wisdom, and to my co-supervisor Lex Nederbragt for valuable comments and insight in bioinformatics. I would also like to thank my good friends Trygve Danielsen for proofreading and Synnøve Smedbye Botnen for sharing her expertise in molecular biology. Thanks to Tom Andersen for inspiration. Thanks to close family for proofreading. Thanks to friends at “Utfallsrommet” for both academic, and not quite so academic, discussions. Last but not least, a very special thanks to Marte Rødsvik Kolloen for unconditional support, encouragement and highly appreciated technical dis- cussions. Martin Gjesdal Bjørndal Blindern, May 2017 v Contents Contents vii 1 Introduction 1 1.1 Objectives of the Thesis . 3 1.2 Genetic Variation . 3 1.3 Graphs for Sequence Data . 4 1.3.1 Motivation for Graph-Representations . 5 1.4 Measuring Distance Between Graphs . 7 1.5 Related Work . 8 1.6 Outline of the Thesis . 9 2 Data 11 2.1 Genomics . 11 2.1.1 Alignment and String Edit Distance . 12 2.1.2 Assembly . 14 2.1.3 Mapping, Resequencing and Variant Detection . 16 2.2 Description of the Data . 17 2.2.1 The Variant Call Format . 18 2.2.2 Reference Sequence . 20 2.3 Variant Graphs . 21 3 Methods 25 3.1 Distance-Based Approaches . 26 3.1.1 Bubble Counter . 26 3.1.2 Graph Edit Distance . 28 3.2 A Model-Based Approach . 31 3.2.1 The Bayes Factor . 31 3.3 Implementation of the Distance Metrics . 37 3.3.1 Bubble Counter and Bayes Factor . 37 3.3.2 Graph Building . 38 3.3.3 Graph Alignment with GEDEVO . 41 3.4 Are the Measures really Metrics? . 43 3.4.1 Bubble Counter . 44 vii viii CONTENTS 3.4.2 Graph Edit Distance . 46 3.4.3 Bayes Factor Measure . 47 3.5 The Permutation Test . 47 3.5.1 Monte Carlo Sampling . 49 3.6 Implementation of the Permutation Test . 50 3.7 Example . 52 4 Results 57 4.1 Distances Between Population Graphs . 57 4.1.1 Bubble Counter . 57 4.1.2 Graph Edit Distance . 58 4.1.3 Bayes Factor . 58 4.2 Visualization by Multidimensional Scaling . 60 4.3 Permutation Tests . 61 4.3.1 p-values . 61 5 Discussion 73 5.1 Distance Metrics . 73 5.1.1 Bubble Counter . 74 5.1.2 Graph Edit Distance . 76 5.1.3 Bayes Factor . 77 5.1.4 Other Distances . 82 5.2 GraphBuilder . 84 5.3 Other Graph Types . 85 5.3.1 Sequence Graph . 85 5.3.2 de Bruijn Graph . 86 5.4 Biological Features . 89 5.4.1 Out of Africa . 90 6 Concluding Remarks 91 6.1 Future Work . 93 A Graph Theory and some Assembly Graphs 95 A.1 Definitions in Graph Theory . 95 A.2 De Bruijn Graphs for Genome Assembly . 97 B Calculations 99 B.1 Conjugate Prior for the Multinomial Distribution . 99 B.2 Calculations for the Bayes Factor in 3.2.1 . 100 B.3 Proof of Proposition 3.1 . 102 CONTENTS ix C Table from Example 3.3 103 D Notation 105 E Glossary 107 F Code 111 F.1 GraphBuilder . 111 F.2 R-script for Calculating Distances . 123 F.3 R-code for Permutation Tests . 127 Bibliography 135 Chapter 1 Introduction DNA data is extensively utilized for discovering differences between organisms, but traditionally only linear representations of sequences has been exploited for this purpose. The idea of our DNA as linear, represented as a string of characters from the alphabet Σ = fA; C; G; T g, prevails (see e.g. Watson, T. Baker, et al. (2008), page 30). However, the question of how to effectively rep- resent collections of DNA sequences has been, and remains, an active subject of discussion (Paten, Novak, and Haussler, 2014; Iqbal, Caccamo, et al., 2012; Church, Schneider, Graves, et al., 2011; Dilthey et al., 2015). This thesis seeks to improve upon the analysis of DNA sequence data by using graphs made up of collections of sequences, instead of the linear representation provided by strings. The aim of the graph-based approach is to include more data in the model, thereby providing more accurate measures of the relative distances between populations. In the following, the word “population” will refer to a group of breeding individuals of the same species (Hartl, 2000; Falconer, 1981). Genetic data has become an extremely powerful tool in medicine and bio- sciences. Facilitated by the collection of vast amounts of data, the possibilities of the combination of biology, mathematics, statistics and computer science are seemingly endless. For example, a milestone of modern science is the completion of the first draft of the human genome (Venter et al., 2001). Full genome sequencing is becoming increasingly common as software is getting more efficient, sequencing technology more accurate, and hardware better. In the mid 2000s, a new group of sequencing technologies emerged. These methods are referred to as High Throughput Sequencing (HTS) and produced massive amounts of data per time compared to earlier designs. The intro- duction of HTS caused a need for specialized data structures and software to handle the large amounts of data. This brought us various graph-based 1 2 CHAPTER 1. INTRODUCTION methods for de novo sequence assembly1 and mapping2 (Limasset and Peter- longo, 2015; Novak, Garrison, and Paten, 2016) (see Section 2.1.2, 2.1.3 and Appendix A.2). Graphs allow us to store variations in the genetic material by using mul- tiple nodes to represent all the possible variations at a position. Hence, the acquisition of several genome sequences from the same population motivates the use of graphs to store the data. A graph made from DNA sequences of several individuals from one single population will in the following be referred to as a population graph. A key concept of this thesis is the comparison of pop- ulation graphs. In order to do calculations on graphs, a Java-program called GraphBuilder (see Subsection 3.3.2) was made.
Load more

Recommended publications
  • Error-Tolerant Graph Matching: a Formal Framework and Algorithms
    Error-Tolerant Graph Matching: A Formal Framework and Algorithms H. Bunke Department of Computer Science,University of Bern, Neubr/ickstr. 10, CH-3012 Bern, Switzerland [email protected] Abstract. This paper first reviews some theoretical results in error- tolerant graph matching that were obtained recently. The results in- clude a new metric for error-tolerant graph matching based on maximum common subgraph, a relation between maximum common subgraph and graph edit distance, and the existence of classes of cost functions for error-tolerant graph matching. Then some new optimal algorithms for error-tolerant graph matching are discussed. Under specific conditions, the new algorithms may be significantly more efficient than traditional methods. Keywords: structural pattern recognition, graphs, graph matching, error-tolerant matching, edit distance, maximum common subgraph, cost function 1 Introduction In structural pattern recognition, graphs are widely used for object representa- tion. Typically, parts of a complex object, or pattern, are represented by nodes, and relations between the various parts by edges. Labels and/or attributes for nodes and edges are used to further distinguish between different classes of nodes and edges, or to incorporate numerical information in the symbolic representa- tion. Using graphs to represent both known models from a database and un- known input patterns, the recognition task turns into a graph matching problem. That is, the database is searched for models that are similar to the unknown input graph. Standard algorithms for graph matching include graph isomorphism, sub- graph isomorphism, and maximum common subgraph search [1, 2]. However, in real world applications we can't always expect a perfect match between the input and one of the graphs in the database.
    [Show full text]
  • Typical Snapshots Selection for Shortest Path Query in Dynamic Road Networks
    Typical Snapshots Selection for Shortest Path Query in Dynamic Road Networks Mengxuan Zhang, Lei Li, Wen Hua, Xiaofang Zhou School of Information Technology and Electrical Engineering The University of Queensland, Brisbane, Australia [email protected], fl.li3, [email protected],[email protected] Abstract. Finding the shortest paths in road network is an important query in our life nowadays, and various index structures are constructed to speed up the query answering. However, these indexes can hardly work in real-life scenario because the traffic condition changes dynamically, which makes the pathfinding slower than in the static environment. In order to speed up path query answering in the dynamic road network, we propose a framework to support these indexes. Firstly, we view the dynamic graph as a series of static snapshots. After that, we propose two kinds of methods to select the typical snapshots. The first kind is time-based and it only considers the temporal information. The second category is the graph representation-based, which considers more insights: edge-based that captures the road continuity, and vertex-based that reflects the region traffic fluctuation. Finally, we propose the snapshot matching to find the most similar typical snapshot for the current traffic condition and use its index to answer the query directly. Extensive experiments on real-life road network and traffic conditions validate the effectiveness of our approach. Keywords: Shortest Path, Snapshot Selection, Dynamic Road Network 1 Introduction Shortest path query is a fundamental operation in road network routing and navigation. A road network can be denoted as a directed graph G(V; E) where V is the set of road intersections, and E ⊆ V × V is the set of road segments.
    [Show full text]
  • On the Unification of the Graph Edit Distance and Graph Matching Problems Romain Raveaux
    On the unification of the graph edit distance and graph matching problems Romain Raveaux To cite this version: Romain Raveaux. On the unification of the graph edit distance and graph matching problems. Pattern Recognition Letters, Elsevier, 2021, 145, 10.1016/j.patrec.2021.02.014. hal-03163084v2 HAL Id: hal-03163084 https://hal.archives-ouvertes.fr/hal-03163084v2 Submitted on 19 Mar 2021 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. On the unification of the graph edit distance and graph matching problems Romain Raveaux1 1Universit´ede Tours, Laboratoire d'Informatique Fondamentale et Appliqu´eede Tours (LIFAT - EA 6300), 64 Avenue Jean Portalis, 37000 Tours, France 7th March 2020 Abstract Error-tolerant graph matching gathers an important family of problems. These problems aim at finding correspondences between two graphs while integrating an error model. In the Graph Edit Distance (GED) problem, the insertion/deletion of edges/nodes from one graph to another is explicitly expressed by the error model. At the opposite, the problem commonly referred to as \graph matching" does not explicitly express such operations. For decades, these two problems have split the research community in two separated parts.
    [Show full text]
  • A Binary Linear Programming Formulation of the Graph Edit Distance
    1200 IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 28, NO. 8, AUGUST 2006 A Binary Linear Programming Formulation of the Graph Edit Distance Derek Justice and Alfred Hero, Fellow, IEEE Abstract—A binary linear programming formulation of the graph edit distance for unweighted, undirected graphs with vertex attributes is derived and applied to a graph recognition problem. A general formulation for editing graphs is used to derive a graph edit distance that is proven to be a metric, provided the cost function for individual edit operations is a metric. Then, a binary linear program is developed for computing this graph edit distance, and polynomial time methods for determining upper and lower bounds on the solution of the binary program are derived by applying solution methods for standard linear programming and the assignment problem. A recognition problem of comparing a sample input graph to a database of known prototype graphs in the context of a chemical information system is presented as an application of the new method. The costs associated with various edit operations are chosen by using a minimum normalized variance criterion applied to pairwise distances between nearest neighbors in the database of prototypes. The new metric is shown to perform quite well in comparison to existing metrics when applied to a database of chemical graphs. Index Terms—Graph algorithms, similarity measures, structural pattern recognition, graphs and networks, linear programming, continuation (homotopy) methods. æ 1INTRODUCTION TTRIBUTED graphs provide convenient structures for chemicals or medicines [6]. Various techniques have been Arepresenting objects when relational properties are of designed for processing the graphs for structural features and interest.
    [Show full text]
  • Bayesian Graph Edit Distance
    This is a repository copy of Bayesian graph edit distance. White Rose Research Online URL for this paper: https://eprints.whiterose.ac.uk/1988/ Article: Myers, R., Hancock, E.R. orcid.org/0000-0003-4496-2028 and Wilson, R.C. (2000) Bayesian graph edit distance. IEEE Transactions on Pattern Analysis and Machine Intelligence. pp. 628-635. ISSN 0162-8828 https://doi.org/10.1109/34.862201 Reuse Items deposited in White Rose Research Online are protected by copyright, with all rights reserved unless indicated otherwise. They may be downloaded and/or printed for private study, or other acts as permitted by national copyright laws. The publisher or other rights holders may allow further reproduction and re-use of the full text version. This is indicated by the licence information on the White Rose Research Online record for the item. Takedown If you consider content in White Rose Research Online to be in breach of UK law, please notify us by emailing [email protected] including the URL of the record and the reason for the withdrawal request. [email protected] https://eprints.whiterose.ac.uk/ 628 IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 22, NO. 6, JUNE 2000 Bayesian Graph Edit Distance strings [10]. This idea has been extended to form a basis for comparing trees and graphs on a global level [11], [12], [1], [7]. Richard Myers, Richard C. Wilson, and More recently, the idea of actively editing graphs during the matching process to eliminate relational clutter has proved very Edwin R. Hancock successful [13], [8], [14].
    [Show full text]
  • A Survey of Graph Edit Distance
    Pattern Anal Applic (2010) 13:113–129 DOI 10.1007/s10044-008-0141-y THEORETICAL ADVANCES A survey of graph edit distance Xinbo Gao Æ Bing Xiao Æ Dacheng Tao Æ Xuelong Li Received: 15 November 2007 / Accepted: 16 October 2008 / Published online: 13 January 2009 Ó Springer-Verlag London Limited 2009 Abstract Inexact graph matching has been one of the 1 Originality and contribution significant research foci in the area of pattern analysis. As an important way to measure the similarity between pair- Graph edit distance is an important way to measure the wise graphs error-tolerantly, graph edit distance (GED) is similarity between pairwise graphs error-tolerantly in the base of inexact graph matching. The research advance inexact graph matching and has been widely applied to of GED is surveyed in order to provide a review of the pattern analysis and recognition. However, there is scarcely existing literatures and offer some insights into the studies any survey of GED algorithms up to now, and therefore of GED. Since graphs may be attributed or non-attributed this paper is novel to a certain extent. The contribution of and the definition of costs for edit operations is various, the this paper focuses on the following aspects: existing GED algorithms are categorized according to these On the basis of authors’ effort for studying almost all two factors and described in detail. After these algorithms GED algorithms, authors firstly expatiate on the idea of are analyzed and their limitations are identified, several GED with illustrations in order to make the explanation promising directions for further research are proposed.
    [Show full text]
  • A Metric Learning Approach to Graph Edit Costs for Regression Linlin Jia, Benoit Gaüzère, Florian Yger, Paul Honeine
    A Metric Learning Approach to Graph Edit Costs for Regression Linlin Jia, Benoit Gaüzère, Florian Yger, Paul Honeine To cite this version: Linlin Jia, Benoit Gaüzère, Florian Yger, Paul Honeine. A Metric Learning Approach to Graph Edit Costs for Regression. Proceedings of IAPR Joint International Workshops on Statistical techniques in Pattern Recognition (SPR 2020) and Structural and Syntactic Pattern Recognition (SSPR 2020), Jan 2021, Venise, Italy. hal-03128664 HAL Id: hal-03128664 https://hal-normandie-univ.archives-ouvertes.fr/hal-03128664 Submitted on 2 Feb 2021 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. A Metric Learning Approach to Graph Edit Costs for Regression Linlin Jia1;4, Benoit Ga¨uz`ere1;4, Florian Yger2?, and Paul Honeine3;4 1 LITIS Lab, INSA Rouen Normandie, France 2 LAMSADE, Universit´eParis Dauphine-PSL, France 3 LITIS Lab, Universit´ede Rouen Normandie, France 4 Normandie Universit´e,France Abstract. Graph edit distance (GED) is a widely used dissimilarity measure between graphs. It is a natural metric for comparing graphs and respects the nature of the underlying space, and provides interpretability for operations on graphs. As a key ingredient of the GED, the choice of edit cost functions has a dramatic effect on the GED and therefore the classification or regression performances.
    [Show full text]
  • Theory of Graph Traversal Edit Distance, Extensions, and Applications
    THESIS THEORY OF GRAPH TRAVERSAL EDIT DISTANCE, EXTENSIONS, AND APPLICATIONS Submitted by Ali Ebrahimpour Boroojeny Department of Computer Science In partial fulfillment of the requirements For the Degree of Master of Science Colorado State University Fort Collins, Colorado Spring 2019 Master’s Committee: Advisor: Hamidreza Chitsaz Asa Ben-Hur Zaid Abdo Copyright by Ali Ebrahimpour Boroojeny 2019 All Rights Reserved ABSTRACT THEORY OF GRAPH TRAVERSAL EDIT DISTANCE, EXTENSIONS, AND APPLICATIONS Many problems in applied machine learning deal with graphs (also called networks), including social networks, security, web data mining, protein function prediction, and genome informatics. The kernel paradigm beautifully decouples the learning algorithm from the underlying geometric space, which renders graph kernels important for the aforementioned applications. In this paper, we give a new graph kernel which we call graph traversal edit distance (GTED). We introduce the GTED problem and give the first polynomial time algorithm for it. Informally, the graph traversal edit distance is the minimum edit distance between two strings formed by the edge labels of respective Eulerian traversals of the two graphs. Also, GTED is motivated by and provides the first mathematical formalism for sequence co-assembly and de novo variation detection in bioinformatics. We demonstrate that GTED admits a polynomial time algorithm using a linear program in the graph product space that is guaranteed to yield an integer solution. To the best of our knowledge, this is the first approach to this problem. We also give a linear programming relaxation algorithm for a lower bound on GTED. We use GTED as a graph kernel and evaluate it by computing the accuracy of an SVM classifier on a few datasets in the literature.
    [Show full text]
  • Graph Edit Networks
    Published as a conference paper at ICLR 2021 GRAPH EDIT NETWORKS Benjamin Paassen Daniele Grattarola The University of Sydney Università della Svizzera italiana [email protected] [email protected] Daniele Zambon Cesare Alippi Università della Svizzera italiana Università della Svizzera italiana [email protected] Politecnico di Milano [email protected] Barbara Hammer Bielefeld University [email protected] ABSTRACT While graph neural networks have made impressive progress in classification and regression, few approaches to date perform time series prediction on graphs, and those that do are mostly limited to edge changes. We suggest that graph edits are a more natural interface for graph-to-graph learning. In particular, graph edits are general enough to describe any graph-to-graph change, not only edge changes; they are sparse, making them easier to understand for humans and more efficient computationally; and they are local, avoiding the need for pooling layers in graph neural networks. In this paper, we propose a novel output layer - the graph edit network - which takes node embeddings as input and generates a sequence of graph edits that transform the input graph to the output graph. We prove that a mapping between the node sets of two graphs is sufficient to construct training data for a graph edit network and that an optimal mapping yields edit scripts that are almost as short as the graph edit distance between the graphs. We further provide a proof-of-concept empirical evaluation on several graph dynamical systems, which are difficult to learn for baselines from the literature.
    [Show full text]
  • Big Graph Processing: Partitioning and Aggregated Querying
    Big Graph Processing : Partitioning and Aggregated Querying Ghizlane Echbarthi To cite this version: Ghizlane Echbarthi. Big Graph Processing : Partitioning and Aggregated Querying. Databases [cs.DB]. Université de Lyon, 2017. English. NNT : 2017LYSE1225. tel-01707153 HAL Id: tel-01707153 https://tel.archives-ouvertes.fr/tel-01707153 Submitted on 12 Feb 2018 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. No d’ordre NNT : 2017LYSE1225 THESEDEDOCTORATDEL’UNIVERSIT` EDELYON´ op´er´ee au sein de l’Universit´e Claude Bernard Lyon 1 Ecole´ Doctorale ED512 Infomath Sp´ecialit´e de doctorat : Informatique Soutenue publiquement le 23/10/2017, par : Ghizlane ECHBARTHI Big Graph Processing: Partitioning and Aggregated Querying Devant le jury compos´ede: Nom Pr´enom, grade/qualit´e, ´etablissement/entreprise Pr´esident(e) Karine Zeitouni, Professeur, Universit´e de Versaille Rapporteure Raphael Couturier, Professeur, Universit´e de Franche-Comt´e Rapporteur Angela Bonifati, Professeur, Universit´e Lyon 1 Examinatrice Mohand Boughanem, Professeur, Universit´e Paul Sabatier Examinateur Hamamache Kheddouci, Professeur, Universit´e Lyon 1 Directeur de th`ese Remerciements Tout d’abord je tiens `a remercier mon directeur de th`ese Hamamache KHEDDOUCI, de m’avoir acceuilli au sein de l’´equipe GOAL, d’avoir constamment veill´e sur la qualit´e de mon travail et de m’avoir apport´e un soutien scientifique et morale d’une grande importance.
    [Show full text]
  • Graph Attribute Embedding Via Riemannian Submersion Learning
    Graph Attribute Embedding via Riemannian Submersion Learning Author Zhao, Haifeng, Robles-Kelly, Antonio, Zhou, Jun, Lu, Jianfeng, Yang, Jing-Yu Published 2011 Journal Title Computer Vision and Image Understanding DOI https://doi.org/10.1016/j.cviu.2010.12.005 Copyright Statement © 2011 Elsevier Inc. This is the author-manuscript version of this paper. Reproduced in accordance with the copyright policy of the publisher. Please refer to the journal's website for access to the definitive, published version. Downloaded from http://hdl.handle.net/10072/52094 Griffith Research Online https://research-repository.griffith.edu.au Graph Attribute Embedding via Riemannian Submersion Learning Haifeng Zhao4 Antonio Robles-Kelly1;2;3 Jun Zhou1;2;3 Jianfeng Lu4 Jing-Yu Yang4 1 NICTA,∗ Locked Bag 8001, Canberra ACT 2601, Australia 2 The Australian National University, Canberra ACT 0200, Australia 3 UNSW@ADFA, Canberra, ACT 2600, Australia 4 Nanjing University of Science and Technology, Nanjing 210094, China Abstract In this paper, we tackle the problem of embedding a set of relational structures into a metric space for purposes of matching and categorisation. To this end, we view the problem from a Riemannian perspective and make use of the concepts of charts on the manifold to define the embedding as a mixture of class-specific submersions. Formulated in this manner, the mixture weights are recovered using a probability density estimation on the embedded graph node coordinates. Further, we recover these class-specific submersions making use of an iterative trust-region method so as to minimise the L2 norm between the hard limit of the graph-vertex posterior probabilities and their estimated values.
    [Show full text]
  • Correlation Constraint Shortest Path Over Large Multi-Relation Graphs
    Correlation Constraint Shortest Path over Large Multi-Relation Graphs ∗ Xiaofei Zhang M. Tamer Ozsu¨ University of Memphis University of Waterloo [email protected] [email protected] ABSTRACT complex protein-to-protein interaction with six types of correla- Multi-relation graphs intuitively capture the heterogeneous corre- tions statistically learned from existing protein databases, revealing lations among real-world entities by allowing multiple types of re- that most protein-to-protein interactions are associated with at least lationships to be represented as entity-connecting edges, i.e., two two types of correlations. Other applications where these multiple entities could be correlated with more than one type of relation- relationships exist between entities include urban and transporta- ship. This is important in various applications such as social net- tion systems [13], ecology research [41], and recommender sys- work analysis, ecology, and bio-informatics. Existing studies on tems [31, 40]. The data in these applications are typically mod- these graphs usually consider an edge label constraint perspective, eled as multi-relation graphs capturing the heterogeneous multiple where each edge contains only one label and each edge is consid- correlations among real-world entities as entity-connecting edges. ered independently. For example, there are lines of research fo- Compared to other emerging graph models, e.g., the property graph cusing on reachability between two vertices under a set of edge model [4, 3], which allows multiple attributes and values to be as- label constraints, or finding paths whose consecutive edge labels sociated with vertices and edges, the multi-relation model is sim- satisfy a user-specified logical expression.
    [Show full text]