Abstract
It has been observed in real networks that the fraction of nodes P(k) with degree k satisfies the power-law P(k) ∝ k−γ for k > kmin > 0. However, the degree distribution of nodes in these networks before kmin varies slowly to the extent of being uniform as compared to the degree distribution after kmin. Most of the previous studies focus on the degree distribution after kmin and ignore the initial flatness in the distribution of degrees. In this paper, we propose a model that describes the degree distribution for the whole range of k > 0, i.e., before and after kmin. The network evolution is made up of two steps. In the first step, a new node is connected to the network through a preferential attachment method. In the second step, a certain number of edges between the existing nodes are added such that the end nodes of an edge are selected either uniformly or preferentially. The model has a parameter to control the uniform or preferential selection of nodes for creating edges in the network. We perform a comprehensive mathematical analysis of our proposed model in the discrete domain and prove that the model exhibits an asymptotically power-law degree distribution after kmin and a flat-ish distribution before kmin. We also develop an algorithm that guides us in determining the model parameters in order to fit the model output to the node degree distribution of a given real network. Our simulation results show that the degree distributions of the graphs generated by this model match well with those of the real-world graphs.
Similar content being viewed by others
References
Ahn YY, Han S, Kwak H, Moon S, Jeong H (2007) Analysis of topological characteristics of huge online social networking services. In: Proceedings of the 16th international conference on world wide web, pp 835–844
Albert R, Barabási AL (2000) Topology of evolving networks: local events and universality. Phys Rev Lett 85:5234–5237
Amaral LAN, Scala A, Barthélémy M, Stanley HE (2000) Classes of small-world networks. Proc Nat Acad Sci 97(21):11149–11152
Anwar R, Yousuf MI, Abid M (2020) Analysis of a model for generating weakly scale-free networks. Discret Math Theoret Comput Sci 22(1):6
Barabási AL, Albert R (1999) Emergence of scaling in random networks. Science 286:509–512
Berger N, Borgs C, Chayes J, D’Souza R, Kleinberg R (2004) Competition-induced preferential attachment. In: International conference on automata, languages and programming
Broder A, Kumar R, Maghoul F, Raghavan P, Rajagopalan S, Stata R, Tomkins A, Wiener J (2000) Graph structure in the web. Comput Netw Int J Comput Telecommun Netw 33:309–320
Chen Y, Ding C, Hu J, Chen R, Hui P, Fu X (2017) Building and analyzing a global co-authorship network using google scholar data. In: Proceedings of 26th international world wide web conference
Collevecchio A, Cotar C, LiCalzi M (2013) On a preferential attachment and generalized pólya’s urn model. Annals Appl Probab 23(3):1219–1253
Cooper C, Frieze A (2003) A general model of web graphs. Random Struct Algor 22(3):311–335
Deijfen M, Esker H, Hofstad R, Hooghiemstra G (2009) A preferential attachment model with random initial degrees. Arkiv fö,r Matematik 47:41–72
Dorogovtsev S, Mendes JF, N Samukhin A (2000) Structure of growing networks with preferential linking. Phys Rev Lett 85:4633–4636
Golder S, Wilkinson D, Huberman B (2006) Rhythms of social interaction: messaging within a massive online network. In: Proceedings of the 3rd international conference on communities and technologies
H Strogatz S (2001) Exploring complex networks. Nat Int J Sci 410:268–276
KONECT (2017) The Koblenz Network Collection - KONECT. http://konect.uni-koblenz.de/networks/
Krapivsky PL, Redner S, Leyvraz F (2000) Connectivity of growing random networks. Phys Rev Lett 85:4629–4632
Kumar R, Novak J, Tomkins A (2006) Structure and evolution of online social networks. In: Proceedings of the 12th ACM SIGKDD international conference on knowledge discovery and data mining, pp 611–617
Lansky P, Polito F, Sacerdote L (2014) The role of detachment of in-links in scale-free networks. J Phys A Math Theoret 47(34):345002
Lansky P, Polito F, Sacerdote L (2016) Generalized nonlinear yule models. J Stat Phys 165(3):661–679
Leskovec J, Krevl A (2014) SNAP Datasets: Stanford Large Network Dataset Collection. http://snap.stanford.edu/data
Mislove A, Marcon M, Gummadi KP, Druschel P, Bhattacharjee B (2007) Measurement and analysis of online social networks. In: Proceedings of the 7th ACM SIGCOMM conference on internet measurement, pp 29–42
Newman MEJ (2001) The structure of scientific collaboration networks. Proc Nat Acad Sci 98(2):404–409
Ostroumova L, Ryabchenko A, Samosvat E (2013) Generalized preferential attachment: tunable power-law degree distribution and clustering coefficient. In: Algorithms and models for the web graph, pp 185–202
Pachón A, Polito F, Sacerdote L (2016) Random graphs associated to some discrete and continuous time preferential attachment models. J Stat Phys 162:1608–1638
Pachón A, Sacerdote L, Yang S (2017) Scale-free behavior of networks with the copresence of preferential and uniform attachment rules. Phys D Nonlinear Phenom 371:1–12
Rossi RA, Ahmed NK (2015) The Network Data Repository with Interactive Graph Analytics and Visualization. http://networkrepositorycom
Rudas A, Tóth B, Valkó B (2007) Random trees and general branching processes. Random Struct Algorit 31(2):186–202
Vázquez A, Pastor-Satorras R, Vespignani A (2002) Large-scale topological and dynamical properties of the internet. Phys Rev E 066130:65
Wang T, Resnick S (2017) Asymptotic normality of in- and out-degree counts in a preferential attachment model. Stoch Model 33:229–255
Zhang P, Mahmoud H (2020) On nodes of small degrees and degree profile in preferential dynamic attachment circuits. Methodol Comput Appl Probab 22:625–645
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix A: Formulation and Derivation of Recurrence Relation for Degree Distribution
Appendix A: Formulation and Derivation of Recurrence Relation for Degree Distribution
In upcoming sections, we will find the exact form of a recurrence relation for the degree distribution and evaluate the relation for the limit \(t \to \infty \).
1.1 A.1 Edge-Step for k = 1
A recurrence relation for k = 1 will be formed and its value for \(t \to \infty \) will be determined in this section.
Rewriting Eq. 9 in the following form,
and repeatedly using the above relation for \(m=1,2,\dots \), we can write N1,t+ 1 in terms of N1,t+ 1,0,
Using the relation given by Eq. 4, we get the following expression for N1,t+ 1,
which on simplification becomes,
Using \(P_{1,t} = \frac {N_{1,t}}{t}\), above relation reduces to
Replacing Δet by α, we get,
Now consider a term occurring on R.H.S., which can be re-written in the form,
This reduces the degree-distribution relationship into the following form,
Applying the limit \(t \to \infty \), we get
which can be written as,
1.2 A.2 Derivation of the Recurrence Relation for Degree Distribution k > 1
An exact form of the recurrence relation will be derived here. The recurrence formula writes Nk, t+ 1, number of nodes of degree k at time t + 1, in terms of Nh, t for \(h=k,k-1, \dots \). The network evolves through both node- and edge- steps. The formulation is valid whether Δet is constant or varies over time t.
We introduce the following functions
and re-write Eq. 8 in the following form forms: for m = 1,
and in general for adding m − th edge,
Define an operator \(\mathcal {D}\) with the following properties,
-
P1:
$$ \mathcal{D}f_{k}= \left\{ \begin{array}{ll} f_{k-1} & \text{if } k > 1 \\ 0 & \text{if } k =1 \end{array} \right. $$ -
P2: \(\mathcal {D}^{n} f_{k}= \mathcal {D}^{n-1} (\mathcal {D}f_{k}) \text { for } n > 1\)
-
P3: \(\mathcal {D} f_{k} h_{k} = (\mathcal {D}f_{k}) (\mathcal {D}h_{k})\)
-
P4: \(\mathcal {D} (f_{k} + h_{k}) = (\mathcal {D}f_{k})+(\mathcal {D}h_{k})\)
Here fk and hk can be either Nk, t, m ,Ψk, m or Φk, m.
Above three relations can be combined to rewrite Nk, t+ 1,3 as,
In the general form, the edge-step gives the following relationship:
The node-step can be written in a modified form,
implying
This is a recurrence relation for k > 1 with \(\mathcal {D}^{k} f_{k} = 0\). Note that product components do not commute due to the occurrence of the operator \(\mathcal {D}\). If we apply the operator \(\mathcal {D}\) to succeeding product components and collect coefficients of Nk−n, t’s, the above recurrence relation takes the following form,
or
Next, we find expressions for these coefficients \(C_{k-n,{{\varDelta }} e_{t}}\)’s,
1.2.1 \({ C_{k, {{\varDelta }} e_{t}}} \)
\(C_{k, {{\varDelta }} e_{t}}\) is the coefficient of Nk, t and can be readily computed,
1.2.2 \({ C_{k-1, {{\varDelta }} e_{t}}} \)
Consider \(C_{k-1,{{\varDelta }} e_{t}}\), coefficient of the Nk− 1,t. The product given by Eq. 20 is expanded into a sum of products. Products in which \(\mathcal {D}\) appears exactly once, form the coefficient of \(C_{k-1,{{\varDelta }} e_{t}}\). The coefficient \(C_{k-1,{{\varDelta }} e_{t}}\) is the sum of two components, say, \(C^{1}_{k-1,{{\varDelta }} e_{t}}\) and \(C^{2}_{k-1,{{\varDelta }} e_{t}}\). This is based on the choice, whether \((1 - \frac {k}{2e_{t}})\) or \(\frac {k-1}{2e_{t}}\mathcal {D}\) is chosen from the product component \((1 - \frac {k}{2e_{t}} + \frac {k-1}{2e_{t}} \mathcal {D})\) appearing in Eq. 20.
The component \(C^{1}_{k-1,{{\varDelta }} e_{t}}\) is given as:
For computing \(C^{2}_{k-1,{{\varDelta }} e_{t}}\), we choose \(\left (1 - \frac {k}{2e_{t}}\right )\) from the binomial \(\left (1 - \frac {k}{2e_{t}} + \frac {k-1}{2e_{t}} \mathcal {D}\right )\). Now, consider the product of binomials, \(\prod \limits _{i={{\varDelta }} e_{t}}^{1}({{\varPsi }}_{k,i} + {{\varPhi }}_{k-1,i}\mathcal {D})\). Exactly one term of the form \({{\varPhi }}_{k-1,i}\mathcal {D}\) is selected. If the term is selected at i − th position, we get
\({{\varPsi }}_{k,{{\varDelta }} e_{t}} {\dots } {{\varPsi }}_{k,i+1} [{{\varPhi }}_{k-1,i}\mathcal {D}] {{\varPsi }}_{k,i-1}\dots {{\varPsi }}_{k,1}\), which briefly can be written in the form,
Operator \(\mathcal {D}\) will apply on each succeeding term Ψk, m (Rule P3).
Now, the chosen position for Φ can vary from 1 to Δet. Resultantly, we get the sum of all such products and find \(C^{2}_{k-1,{{\varDelta }} e_{t}}\),
Eventually, coefficients of Nk− 1,t is
1.2.3 \({ C_{k-2, {{\varDelta }} e_{t}}} \)
Now consider \(C_{k-2,{{\varDelta }} e_{t}}\), the coefficient of Nk− 2,t. As was assumed previously, \(C_{k-2,{{\varDelta }} e_{t}}\) is the sum of two components. For one component, we choose \((1 - \frac {k}{2e_{t}})\) from the the binomial \((1 - \frac {k}{2e_{t}} + \frac {k-1}{2e_{t}} \mathcal {D})\). To get an coefficient in Nk− 2,t, we have to select exactly two terms of the form \({{\varPhi }}_{k-1,i}\mathcal {D}\) from the product of binomials \(\prod \limits _{i={{\varDelta }} e_{t}}^{1} \biggl ({{\varPsi }}_{k,i} + {{\varPhi }}_{k-1,i}\mathcal {D}\biggr )\). If we select these terms at i − th and j − th positions with i > j, we get,
\({{\varPsi }}_{k,{{\varDelta }} e_{t}} {\dots } {{\varPsi }}_{k,i+1} ({{\varPhi }}_{k-1,i}\mathcal {D}) {{\varPsi }}_{k,i-1}{\dots } {\dots } {{\varPsi }}_{k,j+1} ({{\varPhi }}_{k-1,j}\mathcal {D}) {{\varPsi }}_{k,j-1}\dots {{\varPsi }}_{k,1}\)
Operator \(\mathcal {D}\) will apply on each succeeding term Ψk, m and Φk, m (Rule P3). In a brief form,
Summing all such products,
Equivalently,
which becomes
Adding contribution from the second component,
1.2.4 \({C_{k-n,{{\varDelta }} e_{t}}}\) for \(n = 1,2,3,\dots ,{{\varDelta }} e_{t}\)
The general form for \(C_{k-n,{{\varDelta }} e_{t}}\) can be obtained in the similar way. If we select n indices for Φ terms with \( i_{n} > i_{n-1}>{\dots } >i_{1}\), then we get the coefficient,
or in the simplest form,
1.3 A.3 Evaluation of the Recurrence Relation
This section determines the stationary degree distribution. It finds the limiting value (\(\lim \limits t \to \infty \)) of the recurrence relation (21). To accomplish this, we replace Δet by α and evaluate the limit \(\lim \limits _{t \to \infty } t~C_{k-n, {{\varDelta }} e_{t}}\) for \(n=0,1,2,\dots \).
1.3.1 C k, α
Now expanding further and ignoring terms of order \(\frac {1}{t}\) and less,
\({{\varPsi }}_{k,m} = \bigl (1 - \frac {2\delta }{t+1}- \frac {k(1-\delta )}{e_{t}+m} \bigr )\) and \({{\varPhi }}_{k-1,m} = \bigl (\frac {2\delta }{t+1}+ \frac {(k-1)(1-\delta )}{e_{t}+m} \bigr )\)
1.3.2 C k− 1,α
Ck− 1,α is the coefficient of the term Nk− 1,t. It is the sum of two components.
Now expanding further and ignoring terms of order \(\frac {1}{t}\) and less than it,
1.3.3 C k−n, α for n = 2, 3,...α
Ck−n, α has order \(O(\frac {1}{t^{n}})\). Thus tCk−n, α approaches to 0 on applying the limit for n = 2, 3,...α.
Now we have all the limits required to determine the recurrence relation for the degree distribution:
Rights and permissions
About this article
Cite this article
Anwar, R., Yousuf, M.I. & Abid, M. Uniform Preferential Selection Model for Generating Scale-free Networks. Methodol Comput Appl Probab 24, 449–470 (2022). https://doi.org/10.1007/s11009-021-09854-w
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11009-021-09854-w