2015•Unpublished venueRequires access

Transitivity of reciprocal networks

Isabel Fernández, Jorge Finke

Open publisher page 5 citations

Abstract

Network models are a useful tool to describe and predict dynamic relationships in large collections of data. Characterizing these relationships helps us to explain the emergence of structure as a systematic deviation from random connectivity. This paper introduces an event-driven model that captures the effects of three simple network formation mechanisms: random attachment (a generic abstraction of how a new incoming node connects to a network), triad formation (how the new node establishes transitive relationships), and network response (the way the overall network reacts to attachments). Our work focuses on the impact of the latter on clustering and degree distributions. We prove that any initial network will reach stationary local clustering coefficients, and obey an extended power law distribution for the in-degree and an exponential distribution for the out-degree. For the in-degree in particular, the response mechanism amplifies the scaling behavior that results from the other two mechanisms.

About this research paper

What this paper is about

Network models are a useful tool to describe and predict dynamic relationships in large collections of data. Characterizing these relationships helps us to explain the emergence of structure as a systematic deviation from random connectivity. This paper introduces an event-driven model that captures the effects of three simple network formation mechanisms: random attachment (a generic abstraction of how a new incoming node connects to a network), triad formation (how the new node establishes transitive relationships), and network response (the way the overall network reacts to attachments). Our work focuses on the impact of the latter on clustering and degree distributions. We prove that any initial network will reach stationary local clustering coefficients, and obey an extended power law distribution for the in-degree and an exponential distribution for the out-degree. For the in-degree in particular, the response mechanism amplifies the scaling behavior that results from the other two mechanisms.

Why it matters

OpenAlex reports 5 citations for this work. Citation counts describe recorded attention and do not establish research quality.

Key contribution

A contribution statement is not available in the OpenAlex record.

Method / approach

Method details are not available in the OpenAlex metadata.

Main findings

Findings are not separately available in the OpenAlex metadata.

Limitations

Limitations are not available in the OpenAlex metadata.

Applications

Application details are not available in the OpenAlex metadata.

Available abstract

Network models are a useful tool to describe and predict dynamic relationships in large collections of data. Characterizing these relationships helps us to explain the emergence of structure as a systematic deviation from random connectivity. This paper introduces an event-driven model that captures the effects of three simple network formation mechanisms: random attachment (a generic abstraction of how a new incoming node connects to a network), triad formation (how the new node establishes transitive relationships), and network response (the way the overall network reacts to attachments). Our work focuses on the impact of the latter on clustering and degree distributions. We prove that any initial network will reach stationary local clustering coefficients, and obey an extended power law distribution for the in-degree and an exponential distribution for the out-degree. For the in-degree in particular, the response mechanism amplifies the scaling behavior that results from the other two mechanisms.

Key concepts: Transitive relation, Reciprocal, Computer science, Degree distribution, Network formation, Cluster analysis, Degree (music), Theoretical computer science

Related papers

Back to paper searchBrowse research topicsOriginal source
Transitivity of reciprocal networks — Research Paper | ScholarLens