---
_id: '2301783'
abstract:
- lang: eng
text: "This dissertation combines different viewpoints on the emerging \"physics
of networks\". After a general introduction to the new physics of complex networks
(scale-free degree distribution, small worlds, etc.), and ethical considerations
about the protection of data about people, three different aspects of and on complex
networks are presented, each of which formed the topic of a paper.\r\n\r\nIn the
first paper, the network of EU-funded collaborative R&D projects is analyzed by
measuring the \"small world\" effects of small diameter and high clustering. Scale-free
degree distributions are found, and the exponents measured. The results are similiar
to other collaboration networks. Three plausible models (random set, configuration
model similiar to MolloyReed, and an additive version of the latter) are created
and examined to understand the influence of the size distributions of projects
and organisations. While many of the empirical findings are similiar to the \"synthetic
Europes\", the main found difference is higher edge multiplicities of the unimodal
projections in the case of the empirical Europe, implying that previous collaborations
are repeated more often than in the respective random models.\r\n\r\nThe second
paper uses the example of corruption to illustrate how one can extend classical
epidemic processes to generalized epidemic processes (GEP) on complex networks.
There is a strong nonlinear dependence of the transmission probability on the
local density of corruption: To model the resistance of the individual against
corruption, the main neighbour infection process is triggered only above a given
threshold of infected neighbours. The model also includes a global mean-field
influence that simulates the effect of e.g. mass media, and a similiarly global
cleaning process that depends on the ratio of the still uncorrupted nodes. The
local and global processes can exhibit a fatal resonance. In contrast to classical
epidemics, a phase transition in the initial infection ratio is observed, and
studied under variation of the process and the network parameters. A smart transition
finder algorithm is used to avoid critical slowing down in the critical region.
The networks are varied with respect to edge density, triangle density, exponent
of the scale-free degree distribution, and degree correlation (additive, multiplicative).
High clustering accelerates the infection in this model. Scale-free networks with
exponent gamma approaching 2 (resulting in hubs with a very large degree) exhibit
an interesting difference: Such model societies with a multiplicative degree correlation
are easier to completely infect by our corruption process than networks with an
additive degree correlation.\r\n\r\nThe third paper describes a clustering approach
that seems to be new to the physics networks community. Clustering results are
often visualized as block-structured adjacency matrices. When the nodes are clustered
and sorted by their cluster order, the adjacency matrix shows blocks of more-strongly
connected subspaces along the matrix diagonal. The following idea inspired this
new algorithm was: Why not directly sort the nodes into such a block-structure?
An inductively developed deterministic algorithm is described that uses a parameterized
heuristic of mutual distances of all nodes, reorders them by smallest distances
in a linear chain, cuts between clusters at the highest distance jumps, and takes
the one clustering with the best modularity as the end result. The three parameters
influence the mixing of the direct connection weight A_ij, the two-step connections
(A^2)_ij, the N1-neighbourhood similarity, and the N2-neighbourhood similarity
(structural equivalence). A proof-of-concept-implementation suitable for small
networks is described. The algorithmic time complexity is O(N^3) due to the matrix
multiplication; a discussion of possible enhancements to the algorithm is given.
The success of the algorithm is demonstrated through application to several networks:
the Zachary Karate Club, where by this method a 3-clustering could be found that
has a higher-modularity than the clusterings reported in the literature; a set
of 96 tumors that are clustered by their gene-similarity; and clustered topics
of 27000 EU-funded R&D projects."
accept: '1'
author:
- first_name: Andreas
full_name: Krueger, Andreas
last_name: Krueger
bi_doctype: biDissertation
citation:
ama: 'Krueger A. *Structures, processes, and clustering of complex networks*.
Bielefeld (Germany): Bielefeld University; 2007.'
angewandte-chemie: A. Krueger, *Structures, processes, and clustering of complex
networks*, Bielefeld University, Bielefeld (Germany), **2007**.
apa: 'Krueger, A. (2007). *Structures, processes, and clustering of complex
networks*. Bielefeld (Germany): Bielefeld University.'
apa_indent: 'Krueger,
A. (2007). *Structures, processes, and clustering of complex networks*.
Bielefeld (Germany): Bielefeld University.

'
aps: ' A. Krueger, Structures, processes, and clustering of complex networks, (Bielefeld
University, Bielefeld (Germany), 2007).'
bio1: 'Krueger A (2007)

*Structures, processes, and clustering of complex
networks*.

Bielefeld (Germany): Bielefeld University.'
chicago: 'Krueger,
Andreas. 2007. *Structures, processes, and clustering of complex networks*.
Bielefeld (Germany): Bielefeld University.

'
default: 'Krueger A (2007)

Bielefeld (Germany): Bielefeld University.'
dgps: 'Krueger,
A. (2007). *Structures, processes, and clustering of complex networks*.
Bielefeld (Germany): Bielefeld University.

'
frontiers: 'Krueger, A. (2007). Structures, processes, and clustering of complex
networks. Bielefeld (Germany): Bielefeld University.'
harvard1: 'Krueger, A., 2007. *Structures, processes, and clustering of complex
networks*, Bielefeld (Germany): Bielefeld University.'
ieee: ' A. Krueger, *Structures, processes, and clustering of complex networks*, Bielefeld
(Germany): Bielefeld University, 2007.'
lncs: ' Krueger, A.: Structures, processes, and clustering of complex networks.
Bielefeld University, Bielefeld (Germany) (2007).'
mla: 'Krueger, Andreas. *Structures, processes, and clustering of complex networks*.
Bielefeld (Germany): Bielefeld University, 2007.'
wels: 'Krueger, A. (2007): Structures, processes, and clustering of complex networks.
Bielefeld (Germany): Bielefeld University.'
date_created: 2008-02-20T10:45:00Z
date_updated: 2018-07-24T12:58:50Z
ddc:
- '510'
defense_date: 2008-02-05
department:
- _id: '10028'
edit_mode: expert
email: akrueger@physik.uni-bielefeld.de
file:
- access_level: open_access
content_type: application/pdf
date_created: 1970-01-01T00:00:00Z
date_updated: 2017-01-23T10:24:15Z
file_id: '2301786'
file_name: AndreasKrueger_PhD_Networks_v1.00_FINAL_(20071031).pdf
open_access: '1'
relation: main_file
file_date_updated: 2017-01-23T10:24:15Z
first_author: Krueger, Andreas
keyword:
- Netzwerk (Graphentheorie)
- Soziales Netzwerk
- Netzwerkanalyse (Soziologie)
- Korruption
- Netzwerk (Physik)
- Komplexe Netzwerke
- Clustering
- EU-Projekte
- Skalenfreie Degree-Verteilung
- Complex networks
- Scale-free degree distribution
- EU projects
- Corruption
- Clustering
language:
- iso: eng
locked: '1'
oa: 1
pacs_class:
- 89.75.-k
- 02.10.Ox
- 89.65.-s
- 87.23.Ge
- 64.60.Aq
place: Bielefeld (Germany)
publisher: Bielefeld University
status: public
supervisor:
- first_name: Philippe
full_name: Blanchard, Philippe
last_name: Blanchard
supversior_idm: on
title: Structures, processes, and clustering of complex networks
type: bi_dissertation
urn: urn:nbn:de:hbz:361-12478
year: '2007'
...