[Paper Review] A weighted evolving network model more approach to reality

Chuan‐Ji Fu, Qing Ou+5 more|arXiv (Cornell University)|Aug 7, 2004

Complex Network Analysis Techniques3 citations

TL;DR

This paper proposes a weighted evolving network model that incorporates strength-driven preferential attachment and inter-node collaboration among low-strength nodes to better reflect real-world network dynamics. The model successfully reproduces power-law distributions for node degrees, strengths, and edge weights, with explicit droop-head and heavy-tail behavior in degree and strength distributions, enhancing realism over prior models.

ABSTRACT

In search of many social and economical systems, it is found that node strength distribution as well as degree distribution demonstrate the behavior of power-law with droop-head and heavy-tail. We present a new model for the growth of weighted networks considering the connection of nodes with low strengths. Numerical simulations indicate that this network model yields three power-law distributions of the node degrees, node strengths and connection weights. Particularly, the droop-head and heavy-tail effects can be reflected in the first two ones by this new model.

Motivation & Objective

• To address the limitation of existing weighted network models in capturing the droop-head and heavy-tail characteristics observed in real-world networks such as social and collaboration systems.
• To model network evolution with realistic mechanisms, including preferential attachment based on node strength and collaboration among low-strength nodes.
• To reproduce power-law distributions for node degrees, strengths, and edge weights simultaneously, aligning with empirical data from real networks.
• To provide a physically interpretable mechanism for competition and cooperation in evolving networks, particularly in social and economic systems.

Proposed method

• The model begins with an initial network of $N_0$ nodes connected by edges of fixed weight $w_0=1$.
• At each time step, a new node is added and attaches to existing nodes with probability proportional to their current strength $s_i$, implementing strength-driven preferential attachment.
• After each new connection, edge weights are locally adjusted via $\Delta w_{ij} = \delta \frac{w_{ij}}{s_i}$, increasing total strength by $\delta$.
• A threshold $s_c$ identifies low-strength nodes (set $G$), which are allowed to form new edges with each other with probability $p$ to enhance their competitive capacity.
• The process iterates, allowing both new node addition and dynamic reconfiguration among existing nodes.
• The model uses numerical simulations on networks of size $N=6000$ with 20 independent realizations to analyze statistical distributions.

Experimental results

Research questions

• RQ1 How can a weighted evolving network model reproduce the droop-head and heavy-tail features observed in empirical degree and strength distributions?
• RQ2 What mechanisms in network evolution lead to the coexistence of power-law distributions in node degrees, strengths, and edge weights?
• RQ3 How does the inclusion of inter-node collaboration among low-strength nodes affect the structural and statistical properties of the network?
• RQ4 To what extent does strength-driven attachment, rather than degree-driven attachment, better reflect real-world network dynamics?
• RQ5 Can the model replicate the observed power-law behavior in edge weight distributions while preserving realistic features in degree and strength distributions?

Key findings

• The node strength distribution $P(s)$ exhibits a clear power-law decay with droop-head and heavy-tail characteristics, consistent with empirical data from real networks.
• The degree distribution $P(k)$ also follows a power law with droop-head and heavy-tail features, matching observations in real-world networks such as the Internet and collaboration networks.
• The edge weight distribution $P(w)$ follows a relatively strict power law without significant droop-head or heavy-tail deviations, resembling empirical data from real systems.
• For large time $t$, the strength $s_i(t)$ and edge weight $w_j(t)$ of initial nodes increase linearly with time, indicating stable growth dynamics.
• The model successfully reproduces all three power-law distributions—degree, strength, and weight—simultaneously, a key improvement over prior models that failed to capture droop-head and heavy-tail in strength and degree.
• The inclusion of collaboration among low-strength nodes enhances network realism by modeling competitive cooperation, such as in scientific collaborations or small business alliances.