pith. sign in

arxiv: 1906.09081 · v1 · pith:E5ZIU4HHnew · submitted 2019-06-21 · 💻 cs.SI · physics.soc-ph

The Impact of Projection and Backboning on Network Topologies

Michele Coscia , Luca Rossi This is my paper

Pith reviewed 2026-05-25 18:25 UTC · model grok-4.3

classification 💻 cs.SI physics.soc-ph
keywords bipartite networksprojectionbackboningnetwork topologycentralizationunipartite projectionmethod clusteringsocial networks
0
0 comments X

The pith

Combinations of projection and backboning on bipartite networks produce two clusters of unipartite topologies with different centralization.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper examines how different ways of projecting bipartite networks into unipartite ones and then extracting backbones affect the final network structure. Twelve combinations of these methods fall into two groups that create networks with very different topologies. The degree to which the network is centralized also changes depending on which pair of methods is used. This matters because many analyses begin with bipartite data, and the preprocessing steps can determine the properties that are later studied.

Core claim

The twelve projection and backboning method combinations divide into two clusters. One cluster produces unipartite networks with one set of topological features while the other cluster produces networks with a contrasting set of features. The level of centralization in the resulting network depends heavily on the specific combination chosen.

What carries the argument

The grouping of projection-backboning method pairs into two clusters based on the topologies of the unipartite networks they generate.

Load-bearing premise

The clustering of methods and differences in centralization are caused by the projection and backboning combinations rather than by the choice of particular datasets or threshold values.

What would settle it

Applying the twelve method combinations to additional bipartite networks from different domains and observing that they no longer form two distinct clusters or that centralization is unaffected by the combinations.

Figures

Figures reproduced from arXiv: 1906.09081 by Luca Rossi, Michele Coscia.

Figure 1
Figure 1. Figure 1: The toy bipartite network for our examples. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The results of two bipartite projection techniques. Edge [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: (Left) A simplification of DF backboning. The node [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: (Left) The cumulative degree distribution of domains [PITH_FULL_IMAGE:figures/full_fig_p005_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: The edge weight distributions given different network projection (from left to right columns: simple, hyperbolic, ProbS, [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: The topological characteristics (from left to right: share of nodes with at least degree equal to one, clustering coefficient, [PITH_FULL_IMAGE:figures/full_fig_p006_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: The similarities between the networks extracted with different network projection and backbone techniques at different [PITH_FULL_IMAGE:figures/full_fig_p007_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: The levels of centralization (color: green high, red low) for each projection (left to right: simple, hyperbolic, ProbS, [PITH_FULL_IMAGE:figures/full_fig_p008_10.png] view at source ↗
read the original abstract

Bipartite networks are a well known strategy to study a variety of phenomena. The commonly used method to deal with this type of network is to project the bipartite data into a unipartite weighted graph and then using a backboning technique to extract only the meaningful edges. Despite the wide availability of different methods both for projection and backboning, we believe that there has been little attention to the effect that the combination of these two processes has on the data and on the resulting network topology. In this paper we study the effect that the possible combinations of projection and backboning techniques have on a bipartite network. We show that the 12 methods group into two clusters producing unipartite networks with very different topologies. We also show that the resulting level of network centralization is highly affected by the combination of projection and backboning applied.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The manuscript examines combinations of 12 projection and backboning methods applied to bipartite networks. It reports that these combinations form two clusters that produce unipartite networks with substantially different topologies and that the resulting network centralization is strongly dependent on the specific projection-backboning pair chosen.

Significance. If the reported clustering and centralization effects are shown to be robust rather than artifacts of the chosen datasets or fixed thresholds, the work would provide a useful empirical warning about methodological sensitivity in bipartite-to-unipartite conversion, with direct relevance to applied network analysis in social and information networks.

major comments (2)
  1. [Abstract and experimental results] The central claim that the 12 method combinations form two stable clusters with distinct topologies rests on a single collection of bipartite networks and fixed similarity/significance thresholds per method. No systematic variation of thresholds or cross-dataset validation is described, so it remains possible that the observed grouping is driven by the particular data and parameter settings rather than by intrinsic properties of the method combinations (Abstract; experimental results section).
  2. [Abstract] No information is supplied on the number, size, or domain of the bipartite datasets, nor on any statistical tests or controls used to establish the two-cluster grouping or the centralization differences. This absence prevents evaluation of whether the reported effects generalize or are dataset-specific (Abstract).

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on the robustness and clarity of our findings. We address each major comment below and will make targeted revisions to the manuscript.

read point-by-point responses

  1. Referee: [Abstract and experimental results] The central claim that the 12 method combinations form two stable clusters with distinct topologies rests on a single collection of bipartite networks and fixed similarity/significance thresholds per method. No systematic variation of thresholds or cross-dataset validation is described, so it remains possible that the observed grouping is driven by the particular data and parameter settings rather than by intrinsic properties of the method combinations (Abstract; experimental results section).

    Authors: We agree that the study employs fixed thresholds chosen according to each method's standard recommendations and analyzes one collection of networks without systematic threshold sweeps or additional cross-validation sets. The two-cluster structure arises from consistent differences in how the projection-backboning pairs retain or filter edges, as measured by multiple topological statistics. While this does not rule out dataset-specific effects, the grouping aligns with known differences in the underlying statistical models (e.g., significance testing versus simple thresholding). We will revise the experimental results section to explicitly note the fixed-parameter limitation and add a paragraph discussing the scope of the observed clustering. revision: partial

  2. Referee: [Abstract] No information is supplied on the number, size, or domain of the bipartite datasets, nor on any statistical tests or controls used to establish the two-cluster grouping or the centralization differences. This absence prevents evaluation of whether the reported effects generalize or are dataset-specific (Abstract).

    Authors: We will expand the abstract to report the number of bipartite networks examined, their approximate sizes, and application domains. The two-cluster grouping was obtained via hierarchical clustering on a vector of normalized topological metrics (degree distribution, clustering coefficient, centralization, etc.), and centralization differences were quantified by direct computation of the centralization index for each projected network. We will add a brief description of these procedures to the abstract and ensure the methods section already contains the relevant controls. revision: yes

Circularity Check

0 steps flagged

No circularity: direct empirical comparison of method combinations on network data

full rationale

The paper conducts an empirical study applying 12 combinations of projection and backboning techniques to bipartite networks and reports observed differences in resulting unipartite topologies and centralization levels. No derivation chain, equations, fitted parameters presented as predictions, or self-citation load-bearing premises appear in the abstract or described content. Claims rest on direct application to chosen datasets rather than any reduction to inputs by construction. This is the most common honest finding for purely comparative empirical work without mathematical modeling.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No mathematical model, free parameters, or invented entities are introduced; the work is an empirical method-comparison study.

pith-pipeline@v0.9.0 · 5669 in / 929 out tokens · 21270 ms · 2026-05-25T18:25:33.963118+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

27 extracted references · 27 canonical work pages

  1. [1]

    Introduction to the special issue on advances in two-mode social networks

    Filip Agneessens and Martin G Everett. Introduction to the special issue on advances in two-mode social networks. Social Networks, 2(35):145– 147, 2013

    work page 2013

  2. [2]

    Scale-free networks

    Albert-L ´aszl´o Barab´asi and Eric Bonabeau. Scale-free networks. Scien- tific american, 288(5):60–69, 2003

    work page 2003

  3. [3]

    Network backboning with noisy data

    Michele Coscia and Frank MH Neffke. Network backboning with noisy data. In Data Engineering (ICDE), 2017 IEEE 33rd International Conference on, pages 425–436. IEEE, 2017

    work page 2017

  4. [4]

    Deep South: A social anthropological study of caste and class

    Allison Davis, Burleigh Bradford Gardner, and Mary R Gardner. Deep South: A social anthropological study of caste and class . Univ of South Carolina Press, 2009

    work page 2009

  5. [5]

    Does co-authorship lead to higher academic productiv- ity? Oxford Bulletin of Economics and Statistics , 77(3):385–407, 2015

    Lorenzo Ductor. Does co-authorship lead to higher academic productiv- ity? Oxford Bulletin of Economics and Statistics , 77(3):385–407, 2015

    work page 2015

  6. [6]

    Legislative cosponsorship networks in the us house and senate

    James H Fowler. Legislative cosponsorship networks in the us house and senate. Social Networks, 28(4):454–465, 2006

    work page 2006

  7. [7]

    Centrality in social networks conceptual clarification

    Linton C Freeman. Centrality in social networks conceptual clarification. Social networks, 1(3):215–239, 1978

    work page 1978

  8. [8]

    Mapping italian news media political coverage in the lead-up of 2018 general election

    Fabio Giglietto, Laura Iannelli, Luca Rossi, Augusto Valeriani, Nicola Righetti, Francesca Carabini, Giada Marino, Stefano Usai, and Elisabetta Zurovac. Mapping italian news media political coverage in the lead-up of 2018 general election. SSRN, 2018

    work page 2018

  9. [9]

    A network analytic approach to understanding cross- platform audience behavior

    Thomas B Ksiazek. A network analytic approach to understanding cross- platform audience behavior. Journal of Media Economics , 24(4):237– 251, 2011

    work page 2011

  10. [10]

    Uncovering collective listening habits and music genres in bipartite networks

    Renaud Lambiotte and Marcel Ausloos. Uncovering collective listening habits and music genres in bipartite networks. Physical Review E , 72(6):066107, 2005

    work page 2005

  11. [11]

    Biclique communities

    Sune Lehmann, Martin Schwartz, and Lars Kai Hansen. Biclique communities. Physical review E , 78(1):016108, 2008

    work page 2008

  12. [12]

    The backbone structure of audience networks: A new approach to comparing online news consumption across countries

    S ´ılvia Maj ´o-V´azquez, Rasmus K Nielsen, and Sandra Gonz ´alez-Bail´on. The backbone structure of audience networks: A new approach to comparing online news consumption across countries. Political Com- munication, pages 1–14, 2018

    work page 2018

  13. [13]

    What do interlocks do? an analysis, critique, and assessment of research on interlocking directorates

    Mark S Mizruchi. What do interlocks do? an analysis, critique, and assessment of research on interlocking directorates. Annual review of sociology, 22(1):271–298, 1996

    work page 1996

  14. [14]

    The backbone of bipartite projections: Inferring rela- tionships from co-authorship, co-sponsorship, co-attendance and other co-behaviors

    Zachary Neal. The backbone of bipartite projections: Inferring rela- tionships from co-authorship, co-sponsorship, co-attendance and other co-behaviors. Social Networks, 39:84–97, 2014

    work page 2014

  15. [15]

    Scientific collaboration networks

    Mark EJ Newman. Scientific collaboration networks. i. network con- struction and fundamental results. Physical review E , 64(1):016131, 2001

    work page 2001

  16. [16]

    The structure and function of complex networks

    Mark EJ Newman. The structure and function of complex networks. SIAM review, 45(2):167–256, 2003

    work page 2003

  17. [17]

    Modularity and community structure in networks

    Mark EJ Newman. Modularity and community structure in networks. Proceedings of the national academy of sciences , 103(23):8577–8582, 2006

    work page 2006

  18. [18]

    Bipartite graphs in systems biology and medicine: a survey of methods and applications

    Georgios A Pavlopoulos, Panagiota I Kontou, Athanasia Pavlopoulou, Costas Bouyioukos, Evripides Markou, and Pantelis G Bagos. Bipartite graphs in systems biology and medicine: a survey of methods and applications. GigaScience, 7(4):giy014, 2018

    work page 2018

  19. [19]

    The immensely inflated news audience: Assessing bias in self-reported news exposure

    Markus Prior. The immensely inflated news audience: Assessing bias in self-reported news exposure. Public Opinion Quarterly, 73(1):130–143, 2009

    work page 2009

  20. [20]

    Ex- tracting the multiscale backbone of complex weighted networks

    M ´Angeles Serrano, Mari ´an Bogun ´a, and Alessandro Vespignani. Ex- tracting the multiscale backbone of complex weighted networks. Pro- ceedings of the national academy of sciences, 106(16):6483–6488, 2009

    work page 2009

  21. [21]

    How do global audiences take shape? the role of institutions and culture in patterns of web use

    Harsh Taneja and James G Webster. How do global audiences take shape? the role of institutions and culture in patterns of web use. Journal of Communication, 66(1):161–182, 2015

    work page 2015

  22. [22]

    Social network analysis: Methods and applications, volume 8

    Stanley Wasserman and Katherine Faust. Social network analysis: Methods and applications, volume 8. Cambridge university press, 1994

    work page 1994

  23. [23]

    Collective dynamics of small- worldnetworks

    Duncan J Watts and Steven H Strogatz. Collective dynamics of small- worldnetworks. nature, 393(6684):440, 1998

    work page 1998

  24. [24]

    Building and interpreting audience networks: A response to mukerjee, majo-vazquez & gonzalez-bailon

    James G Webster and Harsh Taneja. Building and interpreting audience networks: A response to mukerjee, majo-vazquez & gonzalez-bailon. Journal of Communication , 68(3):E11–E14, 2018

    work page 2018

  25. [25]

    Using random walks to generate associations between objects

    Muhammed A Yildirim and Michele Coscia. Using random walks to generate associations between objects. PloS one, 9(8):e104813, 2014

    work page 2014

  26. [26]

    Solving the apparent diversity- accuracy dilemma of recommender systems

    Tao Zhou, Zolt ´an Kuscsik, Jian-Guo Liu, Mat ´uˇs Medo, Joseph Rush- ton Wakeling, and Yi-Cheng Zhang. Solving the apparent diversity- accuracy dilemma of recommender systems. Proceedings of the National Academy of Sciences , 107(10):4511–4515, 2010

    work page 2010

  27. [27]

    Bipartite network projection and personal recommendation

    Tao Zhou, Jie Ren, Mat ´uˇs Medo, and Yi-Cheng Zhang. Bipartite network projection and personal recommendation. Physical Review E , 76(4):046115, 2007

    work page 2007