Download textbook Mobility Analytics For Spatio Temporal And Social Data First International Workshop Mates 2017 Munich Germany September 1 2017 Revised Selected Papers 1St Edition Christos Doulkeridis ebook all chapter pdf
Download textbook Mobility Analytics For Spatio Temporal And Social Data First International Workshop Mates 2017 Munich Germany September 1 2017 Revised Selected Papers 1St Edition Christos Doulkeridis ebook all chapter pdf
Download textbook Mobility Analytics For Spatio Temporal And Social Data First International Workshop Mates 2017 Munich Germany September 1 2017 Revised Selected Papers 1St Edition Christos Doulkeridis ebook all chapter pdf
https://textbookfull.com/product/spatio-temporal-graph-data-
analytics-1st-edition-venkata-gunturi/
https://textbookfull.com/product/graphical-models-for-
security-4th-international-workshop-gramsec-2017-santa-barbara-
ca-usa-august-21-2017-revised-selected-papers-1st-edition-peng-
liu/
https://textbookfull.com/product/data-management-technologies-
and-applications-6th-international-conference-data-2017-madrid-
spain-july-24-26-2017-revised-selected-papers-joaquim-filipe/
https://textbookfull.com/product/music-technology-with-
swing-13th-international-symposium-cmmr-2017-matosinhos-portugal-
september-25-28-2017-revised-selected-papers-mitsuko-aramaki/
Christos Doulkeridis · George A. Vouros
Qiang Qu · Shuhui Wang (Eds.)
Mobility Analytics
LNCS 10731
for Spatio-Temporal
and Social Data
First International Workshop, MATES 2017
Munich, Germany, September 1, 2017
Revised Selected Papers
123
Lecture Notes in Computer Science 10731
Commenced Publication in 1973
Founding and Former Series Editors:
Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen
Editorial Board
David Hutchison
Lancaster University, Lancaster, UK
Takeo Kanade
Carnegie Mellon University, Pittsburgh, PA, USA
Josef Kittler
University of Surrey, Guildford, UK
Jon M. Kleinberg
Cornell University, Ithaca, NY, USA
Friedemann Mattern
ETH Zurich, Zurich, Switzerland
John C. Mitchell
Stanford University, Stanford, CA, USA
Moni Naor
Weizmann Institute of Science, Rehovot, Israel
C. Pandu Rangan
Indian Institute of Technology, Madras, India
Bernhard Steffen
TU Dortmund University, Dortmund, Germany
Demetri Terzopoulos
University of California, Los Angeles, CA, USA
Doug Tygar
University of California, Berkeley, CA, USA
Gerhard Weikum
Max Planck Institute for Informatics, Saarbrücken, Germany
More information about this series at http://www.springer.com/series/7409
Christos Doulkeridis George A. Vouros
•
Mobility Analytics
for Spatio-Temporal
and Social Data
First International Workshop, MATES 2017
Munich, Germany, September 1, 2017
Revised Selected Papers
123
Editors
Christos Doulkeridis Qiang Qu
University of Piraeus Shenzhen Institutes of Advanced
Piraeus Technology
Greece Shenzhen
China
George A. Vouros
University of Piraeus Shuhui Wang
Piraeus Institute of Computing Technology
Greece Beijing
China
LNCS Sublibrary: SL3 – Information Systems and Applications, incl. Internet/Web, and HCI
This proceedings volume contains revised versions of the papers presented at the First
International Workshop on Mobility Analytics for Spatiotemporal and Social Data
(MATES 2017), held in conjunction with the 43rd International Conference on Very
Large Data Bases (VLDB 2017), in Munich, Germany, on September 1, 2017.
Mobility analytics is a timely topic owing to the ever-increasing number of diverse,
real-life applications, ranging from social media to land, sea, and air surveillance
systems, which produce massive amounts of streaming spatiotemporal data, whose
acquisition, cleaning, representation, aggregation, processing, and analysis pose new
challenges for the data management community. The aim of MATES is to bring
together researchers and practitioners interested in developing data-intensive applica-
tions that analyze big spatiotemporal/societal data, in order to foster the exchange of
new ideas on multidisciplinary real-world problems, propose innovative solutions, and
stimulate further research in the area of big spatiotemporal/societal data management
and analysis. The workshop intends to bridge the gap between researchers and domain
experts, most importantly to raise awareness of real-world problems in critical domains
which require novel data management solutions, tailored to addressing the specific
needs of each domain.
The peer-review process put great emphasis on ensuring a high quality of accepted
contributions. Every paper was reviewed by at least three Program Committee
(PC) members. The MATES PC accepted six submissions (46%) as full papers and
another two submissions (16%) as short papers out of a total of 13 submissions. After
careful revision of accepted papers, based both on comments of reviewers and dis-
cussions during the workshop, the chairs decided to allocate the same maximum page
length for all papers included in this volume.
Apart from the peer-reviewed papers that were presented at the workshop, the
program included two keynote speeches, one from academia and one from the
industrial sector. The first keynote — “Effective and Efficient Community Search” —
was given by Dr. Reynold Cheng, Associate Professor of the Department of Computer
Science at the University of Hong Kong. The second keynote — “How Data Analytics
Enables Advanced AIS Applications” — was given by Ernest Batty, Technical
Director of IMIS Global Limited. After the workshop, both keynote speakers were
invited to submit a paper describing the research objectives and future challenges in
relation with their talks, and these papers are included in this volume.
This volume is structured as follows: In the first part, we include the invited papers
from the keynote speakers. Then, the research papers are grouped in thematic areas:
The second part concerns “Social Networks Analytics and Applications,” while the
third part addresses “Spatiotemporal Mobility Analytics.” In this way, the grouping of
research papers reflects the two major foci of the workshop, namely, mobility analytics
for social data and mobility analytics for spatiotemporal data.
VI Preface
The editors wish to thank the PC members for helping MATES put together a
program of high-quality papers that provides an up-to-date overview of the area of
mobility analytics for spatiotemporal and social data. In addition, the editors would like
to thank all authors for submitting their work to MATES.
On a final note, we wish to mention that this workshop was partially supported by
the European Union’s Horizon 2020 research and innovation programme datAcron:
Big Data Analytics for Time Critical Mobility Forecasting, under grant agreement
number 687591, the CAS Pioneer Hundred Talents Program, and the MOE Key
Laboratory of Machine Perception at Peking University under grant number
K-2017-02.
Program Committee
Natalia Andrienko Fraunhofer Institute IAIS, Germany
Alexander Artikis NCSR Demokritos, Greece
Elena Camossi NATO Centre for Maritime Research
and Experimentation (CMRE), Italy
Christophe Claramunt Naval Academy Research Institute, France
Jose Manuel Cordero Garcia CRIDA, Spain
Christos Doulkeridis University of Piraeus, Greece
Georg Fuchs Fraunhofer Institute IAIS, Germany
Maria Halkidi University of Piraeus, Greece
Anne-Laure Jousselme NATO Centre for Martime Research
and Experimentation (CMRE), Italy
Sofia Karagiorgou University of Piraeus, Greece
Jooyoung Lee Syracuse University, USA
Jiehuan Luo Jinan University, China
Michael Mock Fraunhofer Institute IAIS, Germany
Mohamed Mokbel University of Minnesota, USA
Kjetil Noervaag Norwegian University of Science
and Technology, Norway
Kostas Patroumpas University of Piraeus, Greece
Nikos Pelekis University of Piraeus, Greece
Jiang Qingshan Chinese Academy of Sciences, China
Qiang Qu Shenzhen Institutes of Advanced Technology, China
Cyril Ray Naval Academy Research Institute, France
Giorgos Santipantakis University of Piraeus, Greece
David Scarlatti Boeing Research and Technology Europe, Spain
Liu Siyuan The Pennsylvania State University, USA
Yannis Theodoridis University of Piraeus, Greece
Goce Trajcevski Northwestern University, USA
Akrivi Vlachou University of Piraeus, Greece
George Vouros University of Piraeus, Greece
Shuhui Wang Chinese Academy of Sciences, China
Raymong Wong The Hong Kong University of Science
and Technology, SAR China
Contents
1 Introduction
Due to the developments of gigantic social networks (e.g., Flickr, and Facebook),
the topic of attributed graphs has attracted attention from both industry and
research areas [18–20,29,37,49,53]. Essentially, an attributed graph is a graph,
in which vertices or edges are associated with attributes. The attribute of a vertex
often refers to its features, including its interest, hobbies, and locations, while
the attribute of an edge often indicates the relationships between two vertices.
In this article, we consider two typical kinds of attributed graphs, which are
keyword-based attributed graphs, and spatial-based attributed graphs, and study
the problem of community search on these attributed graphs.
Let us see some examples of attributed graphs. Figure 1 illustrates an
attributed graph, where each vertex represents a social network user, and each
edge represent the friendship between two users. The keywords of each user
describe the interest of that user. Figure 2 depicts a spatial-based attributed
graph with nine users in three cities, and each user has a location. The solid
c Springer International Publishing AG 2018
C. Doulkeridis et al. (Eds.): MATES 2017, LNCS 10731, pp. 1–21, 2018.
https://doi.org/10.1007/978-3-319-73521-4_1
2 Y. Fang and R. Cheng
Tom Jim
Bob
Jack Jason
Eric
Jeff John
Leo
lines represent their social relationship, and the dashed lines denote their
locations.
The problems related to retrieving communities from a graph can generally be
classified into community detection (CD) and community search (CS). In general,
CD methods aim to discovery all communities of a graph [22,38,41,42,44,49,50,
54]. These solutions are not “query-based”, i.e., they are not customized for a
query request (e.g., a user-specified query vertex). Besides, it is not clear how
these algorithms can efficiently return a community that contain a given vertex
q. Moreover, they can take a long time to find all the communities for a large
graph, and so they are not suitable for quick or online retrieval of communities.
To solve issues above, CS methods [1,8,9,32,33,48] have been recently devel-
oped. The CS solutions aim to search the community of a specific query vertex
in an “online” manner, which implies that these approaches are query-based.
However, existing CS algorithms assume non-attributed graphs, and only use
the graph structure information to find communities. Thus, it is desirable to
develop methods for searching communities with the consideration of attributes.
As we will show later, the use of attribute information can significantly improve
the effectiveness of the communities retrieved.
In this article, we systematically study the motivation, applications, features,
technical challenges, algorithms, and experimental evaluation of searching com-
munities over these two kinds of attributed graphs. In the following, we will
detail how to tackle these issues.
On Attributed Community Search 3
particular query user, and the link cohesiveness of the community can also be
controlled. (3) Online evaluation. The SAC search is able to find an SAC from
a large spatial graph quickly once a query request arrives.
Since SACs achieve both high structure and spatial cohesiveness, it can be
applied to many interesting applications including event recommendation (e.g.,
Meetup1 ), social marketing, and geo-social data analysis. For example, Meetup
tracks its users’ mobile phone locations, and suggests interesting location-based
events to them [53]. Suppose that Meetup wishes to recommend an event to a
user u. Then we can first find u’s SAC, whose members are physically close
to u. Events proposed by u’s SAC member v can then be introduced to u, so
that u can meet v if she is interested in v’s activity. Since u’s location changes
constantly, u’s recommendation needs to be updated accordingly.
The SAC search problem is very challenging, because the center and radius
of the smallest MCC containing q are unknown. A basic exact approach takes
O(m×n3 ) time to answer a query, where n and m denote the numbers of vertices
and edges in G. To alleviate this issue, we develop three efficient approximation
algorithms with arbitrary approximation ratio, and an advanced exact algo-
rithm, which is much faster than the basic exact algorithm. We have performed
experiments on real datasets and the results show that our solutions yield better
communities than those produced by existing CS and CD algorithms. Moreover,
the approximation algorithms are much faster than the exact algorithms.
We organize the rest of the article as follows. We review the related works in
Sect. 2. In Sect. 3, we investigate the problem of CS on keyword-based attributed
graphs. In Sect. 4, we examine the problem of CS on spatial-based attributed
graphs. We conclude and discuss the future work in Sect. 5.
1
https://www.meetup.com/.
6 Y. Fang and R. Cheng
2 Related Work
The related works about community retrieval can generally be classified into
community detection (CD) and community search (CS). Table 1 summarizes the
works related to community retrieval. We review them in detail as follows.
vertices being loosely connected with others, but it is more compact in space.
Guo et al. [26] proposed the average linkage (ALK) measure for clustering objects
in spatially constrained graphs. In [12], Expert et al. adapted the modularity
function for spatial networks and proposed a method to uncovered commu-
nities from spatial graphs. In [47], Shakarian et al. modified the well known
Louvain algorithm and used a variant of Newman-Girvan modularity to mine
the geographically dispersed communities from location-based social networks.
In [6], Chen et al. proposed a geo-distance-based method using fast modular-
ity maximization for identifying communities that are both highly topologically
connected and spatially clustered from spatially constrained networks. We will
compare our proposed methods with it in the experiments.
However, these CD algorithms are generally costly and time-consuming, as
they often detect all the communities from an entire network. None of these CD
methods has been shown to be able to quickly detect communities from spatial-
based attributed graphs with millions or billions of vertices. Also, it is not clear
how they can be adapted for online retrieval of communities from large spatial-
based attributed graphs. Thus, it calls for the development of faster algorithms
of performing CS on the spatial-based attributed graphs.
2 3
1 G:{x, y}
A:{w, x, y} Core number Vertices
E:{y, z}
0 J
H:{y, z} J:{x}
B:{x} 1 F, G, H, I
D:{x, y, z}
2 E
I:{x} C:{x, y} 3 A, B, C, D
F:{y}
Observe that k-cores are “nested” [3]: given two positive integers i and j, if
i < j, Hj ⊆ Hi . In Fig. 5(a), H3 is contained in H2 , which is nested in H1 .
2
Without ambiguity, all the attributed graphs mentioned in this section refer to
keyword-based attributed graphs.
On Attributed Community Search 9
A list of core numbers and their respective vertices for Example 1 are shown
in Fig. 5(b). We now formally define the KAC query problem as follows.
Problem 1 (KAC query). Given a graph G(V, E), a positive integer k, a vertex
q ∈ V and a set of keywords S ⊆ W (q), return a set G of graphs, such that
∀Gq ∈ G, the following properties hold:
Among the above steps, the key issue is how to compute Gk [S ]. Since Gk [S ]
should satisfy the structure cohesiveness and keyword cohesiveness. Intuitively,
we have two approaches to compute Gk [S ]: either searching the subgraph satisfy-
ing degree constraint first, followed by further refining with keyword constraints
(called basic-g); or vise versa (called basic-w).4
In this section, we propose a novel index, called CL-tree (Core Label tree),
and keywords into a tree structure. Based on
which organizes both the k-cores
the index, the efficiency of answering KAC query can be improved significantly.
The CL-tree index is built based on the key observation that cores are nested.
must be contained in a k-core.
Specifically, a (k + 1)-core All k-cores
can be
organized into a tree structure5 .
Example 2. Consider the graph in Fig. 5(a). All the k-cores can be organized
into a tree as shown in Fig. 6(a). The height of the tree is 4. For each tree node,
we attach the core number and vertex set of its corresponding k-core.
The tree structure in Fig. 6(a) can be stored compactly, as shown in Fig. 6(b).
The key observation is that, for any internal node p in the tree, the vertex sets
of its child nodes are the subsets of p’s vertex set, because of the inclusion
relationship. To save space cost, we can remove the redundant vertices that are
shared by p’s child nodes from p’s vertex set. After such removal, we obtain
a compressed tree, where each graph vertex appears only once. This structure
3
All the proofs of lemmas in this article can be found in [13].
4
All the pseudocodes of algorithms in this article can be found in [13].
5
We use “node” to mean “CL-tree node” in Sect. 3.
On Attributed Community Search 11
0 0
ABCDE
FGHIJ J x: J
1 1 1 1
x: I
x: G
ABCDEFG HI r1 FG HI y: H
y: F, G
z: H
2 2
y: E
ABCDE r2 E z: E
3 3 w: A
x: A,B,C,D
ABCD r3 ABCD y: A,C,D
z: D
(a) tree structure (b) CL-tree index
constitutes the CL-tree index, the nodes of which are further augmented by
inverted lists (Fig. 6(b)). The space cost of the CL-tree is linear to the size of
G. To summarize, each CL-tree node p has five elements: (1) coreNum: the core
(2) vertexSet: a set of graph vertices; (3) invertedList: a
number of the k-core;
list of <key, value> pairs, where the key is a keyword contained by vertices in
vertexSet and the value is the list of vertices in vertexSet containing key; and
(4) childList: a list of child nodes;
Using the CL-tree, the following two key operations used by our query algo-
rithms (Sect. 3.5), can be performed efficiently.
• Core-locating. Given a vertex q and a core number c, find the k-core with
core number c containing q, by traversing the CL-tree.
• Keyword-checking. Given a k-core, find vertices which contain a given
query keyword set, by intersecting the inverted lists of query keywords.
numbers are created prior to those with smaller core numbers. More detailed
steps and analysis of basic and advanced are described in [16].
3.6 Experiments
We consider four real keyword-based attributed graphs. For each of them, each
vertex has a list of neighbors as well as a set of keywords. More details of these
graphs are described in [16]. To evaluate KAC queries, we set the default value
of k to 6. The input keyword set S is set to the whole set of keywords contained
by the query vertex. For each dataset, we randomly select 300 query vertices
with core numbers of 6 or more, which ensures that there is a k-core containing
each query vertex.
We have performed extensive experiments on these datasets. The detailed
experimental results can be found in [16]. The general conclusions observed from
the experiments are that: (1) The communities returned by KAC queries achieve
higher keyword cohesiveness than the state-of-the-art CD and CS methods. For
example, the Jaccard similarities of members in the KACs are higher than those
of Global [48], Local [9], and CODICIL [44]. (2) The case studies on the DBLP
network show that, using keywords, KAC query can find more meaningful com-
munities than Global and Local. (3) The index-based query algorithms are over
1 to 3 orders of magnitude faster than the basic methods. For example, on the
largest dataset DBpedia, a single KAC query takes less than 1 s. (4) For the
index construction methods, advanced is much faster than basic.
On Attributed Community Search 13
y I I
1
6 H H
A F A 2 F
G G
4 3
D D
Q Q
2 B C E
B C E
0
0 2 4 6 8 x
(a) spatial graph (b) k-core decomposition
6
For simplicity, in this section we call spatial-based attributed graphs spatial graphs.
14 Y. Fang and R. Cheng
We first present a basic exact algorithm Exact, which takes O(m×n3 ) to answer
a single query. This is very time-consuming for large graphs. So we turn to design
more efficient approximation algorithms. Here, the approximation ratio is defined
as the ratio of the radius of MCC returned over that of the optimal solution.
Inspired by the approximation algorithms, we also design a fast exact algorithm
Exact+. Their approximation ratios and time complexities are summarized in
Table 2, where F and A are parameters specified by the query user. The value
|F1 | is the number of “fixed vertices”, which will be defined later (|F1 | n).
AppInc is a 2-approximation algorithm, and it is much faster than Exact.
Inspired by AppInc, we design another (2 + F )-approximation algorithm
AppFast, where F ≥ 0, which is faster than AppInc. The limitation of AppInc
Lemma 2 [11]. Given a set S (|S| ≥ 2) of vertices, its MCC can be determined
by at most three vertices in S which lie on the boundary of the circle. If it
is determined by only two vertices, then the line segment connecting those two
vertices must be a diameter of the circle. If it is determined by three vertices,
then the triangle consisting of those three vertices is not obtuse.
By Lemma 2, there are at least two or three vertices lying on the boundary of
the MCC of the target SAC. We call vertices lying on the boundary of an MCC
fixed vertices. So a straightforward method of SAC search can follow the two-
containing q, which is the same
step framework directly. It first finds the k-core
as Global does, and then returns the subgraph achieving both the structure and
spatial cohesiveness by enumerating all the combinations of three vertices in the
We call this method Exact. It completes in O(m × n3 ) time.
k-core.
Lemma 3. 1
2δ ≤ ropt ≤ γ.
Lemma 4. The radius of the MCC covering the feasible solution Φ has an
approximation ratio of 2.
Corollary 1. The center point, o, of the MCC O(o, ropt ) covering Ψ is in the
circle O(q, γ).
ropt
y
6
4 o
q c
rmin
2
0
0 2 4 6 8 x
(a) Splitting O(q, γ) (b) rmin
References
1. Barbieri, N., Bonchi, F., Galimberti, E., Gullo, F.: Efficient and effective commu-
nity search. DMKD 29(5), 1406–1433 (2015)
2. Barthélemy, M.: Spatial networks. Phys. Rep. 499(1), 1–101 (2011)
3. Batagelj, V., Zaversnik, M.: An o(m) algorithm for cores decomposition of net-
works. arXiv (2003)
4. Blei, D.M., Ng, A.Y., Jordan, M.I.: Latent Dirichlet allocation. J. Mach. Learn.
Res. 3, 993–1022 (2003)
5. Bollobás, B.: The evolution of random graphs. Trans. Am. Math. Soc. 286(1),
257–274 (1984)
6. Chen, Y., Jun, X., Minzheng, X.: Finding community structure in spatially con-
strained complex networks. IJGIS 29(6), 889–911 (2015)
7. Cohen, J.: Trusses: cohesive subgraphs for social network analysis. National Secu-
rity Agency Technical Report, p. 16 (2008)
8. Cui, W., Xiao, Y., Wang, H., Lu, Y., Wang, W.: Online search of overlapping
communities. In: SIGMOD, pp. 277–288 (2013)
9. Cui, W., Xiao, Y., Wang, H., Wang, W.: Local search of communities in large
graphs. In: SIGMOD, pp. 991–1002 (2014)
10. Elzinga, D.J., Hearn, D.W.: The minimum covering sphere problem. Manage. Sci.
19(1), 96–104 (1972)
11. Elzinga, D.J., Hearn, D.W.: Geometrical solutions for some minimax location prob-
lems. Transp. Sci. 6(4), 379–394 (1972)
12. Expert, P., et al.: Uncovering space-independent communities in spatial networks.
PNAS 108(19), 7663–7668 (2011)
13. Fang, Y.: Effective and efficient community search over large attributed graphs.
HKU Ph.D. thesis, September 2017
14. Fang, Y., Cheng, R., Chen, Y., Luo, S., Hu, J.: Effective and efficient attributed
community search. VLDB J. 26(6), 803–828 (2017)
15. Fang, Y., Cheng, R., Li, X., Luo, S., Hu, J.: Effective community search over large
spatial graphs. PVLDB 10(6), 709–720 (2017)
16. Fang, Y., Cheng, R., Luo, S., Hu, J.: Effective community search for large
attributed graphs. PVLDB 9(12), 1233–1244 (2016)
17. Fang, Y., Cheng, R., Luo, S., Hu, J., Huang, K.: C-explorer: browsing communities
in large graphs. PVLDB 10(12), 1885–1888 (2017)
18. Fang, Y., Cheng, R., Tang, W., Maniu, S., Yang, X.: Scalable algorithms for
nearest-neighbor joins on big trajectory data. TKDE 28(3), 785–800 (2016)
19. Fang, Y., Cheng, R., Tang, W., Maniu, S., Yang, X.S.: Scalable algorithms for
nearest-neighbor joins on big trajectory data. In: ICDE, pp. 1528–1529 (2016)
20. Fang, Y., Zhang, H., Ye, Y., Li, X.: Detecting hot topics from Twitter: a multiview
approach. J. Inf. Sci. 40(5), 578–593 (2014)
21. Finkel, R.A., Bentley, J.L.: Quad trees: a data structure for retrieval on composite
keys. Acta Informatica 4(1), 1–9 (1974)
22. Fortunato, S.: Community detection in graphs. Phys. Rep. 486(3), 75–174 (2010)
20 Y. Fang and R. Cheng
23. Gaertler, M., Patrignani, M.: Dynamic analysis of the autonomous system graph.
In: IPS, pp. 13–24 (2004)
24. Gibbons, A.: Algorithmic Graph Theory. Cambridge University Press, Cambridge
(1985)
25. Girvan, M., Newman, M.E.J.: Community structure in social and biological net-
works. PNAS 99(12), 7821–7826 (2002)
26. Guo, D.: Regionalization with dynamically constrained agglomerative clustering
and partitioning (redcap). IJGIS 22(7), 801–823 (2008)
27. Guo, T., Cao, X., Cong, G.: Efficient algorithms for answering the m-closest key-
words query. In: SIGMOD, pp. 405–418. ACM (2015)
28. Han, J., Pei, J., Yin, Y.: Mining frequent patterns without candidate generation.
In: SIGMOD (2000)
29. Hu, J., Cheng, R., Huang, Z., Fang, Y., Luo, S.: On embedding uncertain graphs.
In: CIKM. ACM (2017)
30. Hu, J., Wu, X., Cheng, R., Luo, S., Fang, Y.: Querying minimal Steiner maximum-
connected subgraphs in large graphs. In: CIKM, pp. 1241–1250 (2016)
31. Hu, J., Xiaowei, W., Cheng, R., Luo, S., Fang, Y.: On minimal steiner maximum-
connected subgraph queries. TKDE 29(11), 2455–2469 (2017)
32. Huang, X., Cheng, H., Qin, L., Tian, W., Yu, J.X.: Querying k-truss community
in large and dynamic graphs. In: SIGMOD (2014)
33. Huang, X., Lakshmanan, L.V.S., Yu, J.X., Cheng, H.: Approximate closest com-
munity search in networks. PVLDB 9(4), 276–287 (2015)
34. Kim, Y., Son, S.-W., Jeong, H.: Finding communities in directed networks. Phys.
Rev. E 81(1), 016103 (2010)
35. Leicht, E.A., Newman, M.E.J.: Community structure in directed networks. Phys.
Rev. Lett. 100(11), 118703 (2008)
36. Li, R.-H., Qin, L., Yu, J.X., Mao, R.: Influential community search in large net-
works. In: PVLDB (2015)
37. Li, Z., Fang, Y., Liu, Q., Cheng, J., Cheng, R., Lui, J.: Walking in the cloud:
parallel simrank at scale. PVLDB 9(1), 24–35 (2015)
38. Liu, Y., Niculescu-Mizil, A., Gryc, W.: Topic-link LDA: joint models of topic and
author community. In: ICML (2009)
39. Malliaros, F.D., Vazirgiannis, M.: Clustering and community detection in directed
networks: a survey. Phys. Rep. 533(4), 95–142 (2013)
40. Megiddo, N.: Linear-time algorithms for linear programming in r3 and related
problems. In: FOCS, pp. 329–338. IEEE (1982)
41. Nallapati, R.M., Ahmed, A., Xing, E.P., Cohen, W.W.: Joint latent topic models
for text and citations. In: KDD (2008)
42. Newman, M.E.J., Girvan, M.: Finding and evaluating community structure in net-
works. Phys. Rev. E 69(2), 026113 (2004)
43. Plantié, M., Crampes, M.: Survey on social community detection. In: Ramzan,
N., van Zwol, R., Lee, J.S., Clüver, K., Hua, X.S. (eds.) Social Media Retrieval.
Computer Communications and Networks, pp. 65–85. Springer, London (2013).
https://doi.org/10.1007/978-1-4471-4555-4 4
44. Ruan, Y., Fuhry, D., Parthasarathy, S.: Efficient community detection in large
networks using content and links. In: WWW (2013)
45. Sachan, M., et al.: Using content and interactions for discovering communities in
social networks. In: WWW, pp. 331–340 (2012)
46. Seidman, S.B.: Network structure and minimum degree. Soc. Netw. 5(3), 269–287
(1983)
Another random document with
no related content on Scribd:
quite impervious, but becoming—at any rate in the case of the larger
and more important pair—open previous to the final ecdysis. We
have mentioned the contradictory opinions of Réaumur and Dufour,
and will now add the views of some modern investigators. Oustalet
says[341] that there are two pairs of spiracles in the nymphs; the first
pair is quite visible to the naked eye, and is situate between pro- and
meso-notum; it is in the nymph closed by a membrane. The other
pair of spiracles is placed above the posterior pair of legs, is small
and completely closed. He does not state what stage of growth was
attained by the nymphs he examined. Palmén was of opinion that
not only thoracic but abdominal spiracles exist in the nymph,[342] and
that they are completely closed so that no air enters them; he says
that the spiracles have tracheae connected with them, that at each
moult the part closing the spiracles is shed with some of the tracheal
exuviae attached to it. The breathing orifices are therefore for a short
time at each ecdysis open, being subsequently again closed by
some exudation or secretion. This view of Palmén's has been
thought improbable by Hagen and Dewitz, who operated by placing
nymphs in alcohol or warm water and observing the escape of
bubbles from the spots where the supposed breathing orifices are
situate. Both these observers found much difference in the results
obtained in the cases of young and of old nymphs. Hagen concludes
that the first pair of thoracic spiracles are functionally active, and that
abdominal stigmata exist though functionless; he appears to be of
opinion that when the first thoracic stigma is closed this is the result
of the abutting against it of a closed trachea. Dewitz found[343] that
in the adult nymph of Aeschna the thoracic stigma is well developed,
while the other stigmata—to what number and in what position is not
stated—are very small. In a half-grown Aeschnid nymph he found
the thoracic stigma to be present in an undeveloped form. On
placing a full-grown nymph in alcohol, gas escaped from the stigma
in question, but in immature nymphs no escape of gas occurred
although they were subjected to a severe test. A specimen that,
when submitted to the above-mentioned immersion, emitted gas,
subsequently moulted, and thereafter air escaped from the spiracle
previously impervious. The observations of Hagen and Dewitz are
perhaps not so adverse to the views of Palmén as has been
supposed, so that it would not be a matter for surprise if Palmén's
views on this point should be shown to be quite correct.
Odonata are among the few kinds of Insects that are known to form
swarms and migrate. Swarms of this kind have been frequently
observed in Europe and in North America; they usually consist of
species of the genus Libellula, but species of various other genera
also swarm, and sometimes a swarm may consist of more than one
species. L. quadrimaculata is the species that perhaps most
frequently forms these swarms in Europe; a large migration of this
species is said to occur every year in the Charente inférieure from
north to south.[346] It is needless to say that the instincts and stimuli
connected with these migrations are not understood.
The Odonata have no close relations with any other group of Insects.
They were associated by Latreille with the Ephemeridae, in a family
called Subulicornia. The members of the two groups have, in fact, a
certain resemblance in some of the features of their lives, especially
in the sudden change, without intermediate condition, from aquatic to
aerial life; but in all important points of structure, and in their
dispositions, dragon-flies and may-flies are totally dissimilar, and
there is no intermediate group to connect them. We have already,
said that the Odonata consist of two very distinct divisions—
Anisopterides and Zygopterides. The former group comprises the
subfamilies Gomphinae, Cordulegasterinae, Aeschninae,
Corduliinae, and Libellulinae,—Insects having the hinder wings
slightly larger than the anterior pair; while the Zygopterides consist of
only two subfamilies—Calepteryginae and Agrioninae; they have the
wings of the two pairs equal in size, or the hinder a little the smaller.
The two groups Gomphinae and Calepteryginae are each, in several
respects, of lower development than the others, and authorities are
divided in opinion as to which of the two should be considered the
more primitive. It is therefore of much interest to find that there exists
an Insect that shares the characters of the two primitive subfamilies
in a striking manner. This Insect, Palaeophlebia superstes (Fig. 272),
has recently been discovered in Japan, and is perhaps the most
interesting dragon-fly yet obtained. De Selys Longchamps refers it to
the subfamily Calepteryginae, on account of the nature of its wings;
were the Insect, however, deprived of these organs, no one would
think of referring Palaeophlebia to the group in question, for it has
the form, colour, and appearance of a Gomphine Odonate.
Moreover, the two sexes differ in an important character,—the form
of the head and eyes. In this respect the female resembles a
Gomphine of inferior development; while the male, by the shape and
large size of the ocular organs, may be considered to combine the
characters of Gomphinae and Calepteryginae. The Insect is very
remarkable in colour, the large eyes being red in the dead examples.
We do not, however, know what may be their colour during life, as
only one pair of the species is known, and there is no record as to
the life-history and habits. De Selys considers the nearest ally of this
Insect to be Heterophlebia dislocata, a fossil dragon-fly found in the
Lower Lias of England.
CHAPTER XIX
Fig. 281.—A, Last three abdominal segments and bases of the three
caudal processes of Cloëon dipterum: r, dorsal vessel; kl, ostia
thereof; k, special terminal chamber of the dorsal vessel with its
entrance a; b, blood-vessel of the left caudal process; B, twenty-
sixth joint of the left caudal process from below; b, a portion of the
blood-vessel; o, orifice in the latter. (After Zimmermann.)
The life-history has not been fully ascertained in the case of any
species of may-fly; it is known, however, that the development of the
nymph sometimes occupies a considerable period, and it is thought
that in the case of some species this extends to as much as three
years. It is rare to find the post-embryonic development of an Insect
occupying so long a period, so that we are justified in saying that
brief as may be the life of the may-fly itself, the period of preparation
for it is longer than usual. Réaumur says, speaking of the winged fly,
that its life is so short that some species never see the sun. Their
emergence from the nymph-skin taking place at sunset, the duties of
the generation have been, so far as these individuals are concerned,
completed before the morning, and they die before sunrise. He
thinks, indeed, that individuals living thus long are to be looked on as
Methuselahs among their fellows, most of whom, he says, live only
an hour or half an hour.[364] It is by no means clear to which species
these remarks of Réaumur refer; they are doubtless correct in
certain cases, but in others the life of the adult is not so very short,
and in some species may, in all probability, extend over three or four
days; indeed, if the weather undergo an unfavourable change so as
to keep them motionless, the life of the flies may be prolonged for a
fortnight.
Nearly 300 species of Ephemeridae are known, but this may be only
a fragment of what actually exist, very little being known of may-flies