[Paper Review] Uncovering the Temporal Dynamics of Diffusion Networks

Manuel Gomez-Rodriguez, David Balduzzi+1 more|arXiv (Cornell University)|May 3, 2011

Complex Network Analysis Techniques18 references380 citations

TL;DR

This paper proposes NetRate, a convex optimization framework that infers the temporal dynamics of diffusion networks by modeling transmission rates as continuous, edge-specific processes. It uniquely estimates both network structure and time-varying transmission rates from cascade data, achieving high precision and recall without parameter tuning or heuristic sparsity penalties, outperforming prior methods on synthetic and real-world data including hyperlink cascades.

ABSTRACT

Time plays an essential role in the diffusion of information, influence and disease over networks. In many cases we only observe when a node copies information, makes a decision or becomes infected -- but the connectivity, transmission rates between nodes and transmission sources are unknown. Inferring the underlying dynamics is of outstanding interest since it enables forecasting, influencing and retarding infections, broadly construed. To this end, we model diffusion processes as discrete networks of continuous temporal processes occurring at different rates. Given cascade data -- observed infection times of nodes -- we infer the edges of the global diffusion network and estimate the transmission rates of each edge that best explain the observed data. The optimization problem is convex. The model naturally (without heuristics) imposes sparse solutions and requires no parameter tuning. The problem decouples into a collection of independent smaller problems, thus scaling easily to networks on the order of hundreds of thousands of nodes. Experiments on real and synthetic data show that our algorithm both recovers the edges of diffusion networks and accurately estimates their transmission rates from cascade data.

Motivation & Objective

• To infer the underlying network structure and transmission rates of diffusion processes from only observed infection times.
• To model temporal dynamics of information, influence, or disease spread as continuous, edge-specific processes with varying transmission rates.
• To develop a method that naturally enforces sparsity and avoids heuristic parameter tuning.
• To scale efficiently to large networks with hundreds of thousands of nodes through decoupled optimization.
• To improve upon prior methods by estimating continuous transmission rates rather than assuming uniform or fixed rates across edges.

Proposed method

• The method models diffusion as a network of continuous-time, conditionally independent processes with edge-specific transmission rate functions.
• It formulates the inference problem as a convex maximum-likelihood optimization that jointly estimates network edges and transmission rates.
• The likelihood function is based on survival analysis, using log-survival functions and hazard functions to model infection timing.
• The optimization problem decouples into independent subproblems per node, enabling efficient parallelization and scalability.
• The model uses parametric forms (e.g., exponential) for transmission likelihoods, with parameters inferred directly from cascade data.
• Sparsity is naturally induced by the probabilistic model, eliminating the need for heuristic $l_1$-like penalties.

Experimental results

Research questions

• RQ1 Can we infer the underlying network structure and time-varying transmission rates from only observed infection times in a diffusion cascade?
• RQ2 How can we model heterogeneous transmission dynamics across different edges in a diffusion network?
• RQ3 Can we achieve scalable, convex inference without requiring manual parameter tuning for sparsity?
• RQ4 How does the performance of our method compare to state-of-the-art approaches that assume uniform transmission rates or use heuristic sparsity controls?
• RQ5 To what extent can the model recover true network structure and transmission rates on real-world hyperlink and synthetic cascades?

Key findings

• NetRate achieves high precision and recall on synthetic networks with heterogeneous temporal dynamics, outperforming NetInf and ConNIe across a range of parameters.
• On real hyperlink data, NetRate outperforms NetInf and ConNIe in accuracy across the full range of their tunable parameters, despite using a simple parametric model.
• The method produces a unique solution, unlike NetInf and ConNIe, which generate solution curves requiring arbitrary selection.
• NetRate accurately estimates transmission rates, a capability absent in previous methods like NetInf and ConNIe.
• The model's convex formulation enables efficient scaling to networks with hundreds of thousands of nodes through natural parallelization.
• The absence of heuristic sparsity penalties leads to better performance than ConNIe, which relies on such terms and shows degraded results on heterogeneous dynamics.