Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TDNetGen: Empowering Complex Network Resilience Prediction with Generative Augmentation of Topology and Dynamics

Chang Liu, Jingtao Ding, Yiwen Song, Yong Li

TL;DR

TDNetGen tackles resilient prediction for complex networks under label scarcity by learning from unlabeled data through generative augmentation of both topology and nodal dynamics. It decouples the topology distribution diffusion from neural-ODE-based dynamics learning and fuses their outputs in a Transformer+GCN resilience predictor, with classifier-guided diffusion to steer generation toward desired resilience. Empirical results across three network domains show substantial gains in F1 and ACC compared to baselines, and the approach remains effective under limited labels and partial trajectory information, underscoring robustness and practical utility. By enabling data-driven resilience analysis without strong dynamical priors, TDNetGen offers a scalable, reproducible framework with potential impact on the design and protection of real-world complex networks.

Abstract

Predicting the resilience of complex networks, which represents the ability to retain fundamental functionality amidst external perturbations or internal failures, plays a critical role in understanding and improving real-world complex systems. Traditional theoretical approaches grounded in nonlinear dynamical systems rely on prior knowledge of network dynamics. On the other hand, data-driven approaches frequently encounter the challenge of insufficient labeled data, a predicament commonly observed in real-world scenarios. In this paper, we introduce a novel resilience prediction framework for complex networks, designed to tackle this issue through generative data augmentation of network topology and dynamics. The core idea is the strategic utilization of the inherent joint distribution present in unlabeled network data, facilitating the learning process of the resilience predictor by illuminating the relationship between network topology and dynamics. Experiment results on three network datasets demonstrate that our proposed framework TDNetGen can achieve high prediction accuracy up to 85%-95%. Furthermore, the framework still demonstrates a pronounced augmentation capability in extreme low-data regimes, thereby underscoring its utility and robustness in enhancing the prediction of network resilience. We have open-sourced our code in the following link, https://github.com/tsinghua-fib-lab/TDNetGen.

TDNetGen: Empowering Complex Network Resilience Prediction with Generative Augmentation of Topology and Dynamics

TL;DR

TDNetGen tackles resilient prediction for complex networks under label scarcity by learning from unlabeled data through generative augmentation of both topology and nodal dynamics. It decouples the topology distribution diffusion from neural-ODE-based dynamics learning and fuses their outputs in a Transformer+GCN resilience predictor, with classifier-guided diffusion to steer generation toward desired resilience. Empirical results across three network domains show substantial gains in F1 and ACC compared to baselines, and the approach remains effective under limited labels and partial trajectory information, underscoring robustness and practical utility. By enabling data-driven resilience analysis without strong dynamical priors, TDNetGen offers a scalable, reproducible framework with potential impact on the design and protection of real-world complex networks.

Abstract

Predicting the resilience of complex networks, which represents the ability to retain fundamental functionality amidst external perturbations or internal failures, plays a critical role in understanding and improving real-world complex systems. Traditional theoretical approaches grounded in nonlinear dynamical systems rely on prior knowledge of network dynamics. On the other hand, data-driven approaches frequently encounter the challenge of insufficient labeled data, a predicament commonly observed in real-world scenarios. In this paper, we introduce a novel resilience prediction framework for complex networks, designed to tackle this issue through generative data augmentation of network topology and dynamics. The core idea is the strategic utilization of the inherent joint distribution present in unlabeled network data, facilitating the learning process of the resilience predictor by illuminating the relationship between network topology and dynamics. Experiment results on three network datasets demonstrate that our proposed framework TDNetGen can achieve high prediction accuracy up to 85%-95%. Furthermore, the framework still demonstrates a pronounced augmentation capability in extreme low-data regimes, thereby underscoring its utility and robustness in enhancing the prediction of network resilience. We have open-sourced our code in the following link, https://github.com/tsinghua-fib-lab/TDNetGen.
Paper Structure (41 sections, 1 theorem, 27 equations, 8 figures, 3 tables)

This paper contains 41 sections, 1 theorem, 27 equations, 8 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Resilience Prediction
  4. Problem Formulation
  5. Methodology
  6. Overview of Proposed Framework
  7. Topology Diffusion Module
  8. Dynamics Learning Module
  9. Resilience Predictor
  10. Joint Data Augmentation of Topology and Dynamics
  11. Time Complexity Analysis
  12. Experiments
  13. Experimental Settings
  14. Dataset
  15. Baselines and metrics
  16. ...and 26 more sections

Key Result

lemma 1

Denote the forward process conditioned on $y_G$ as $\hat{q}$, and the unconditional forward process as $q$. Given the reasonable assumption $\hat{q}(G^s|G, y_G) = q(G^s|G)$, we have

Figures (8)

  • Figure 1: Resilience of complex networks. $\langle \mathbf{x}\rangle$ denotes the averaged nodal state of the network.
  • Figure 2: Overview of the proposed framework TDNetGen.
  • Figure 3: Illustration of topology diffusion module.
  • Figure 4: Architecture of the resilience predictor.
  • Figure 5: Ablation studies on datasets. CG: Classifier guidance, Diff: Diffusion module, FT: Fine-tuning on dynamics learning module-produced trajectories.
  • ...and 3 more figures

Theorems & Definitions (1)

  • lemma 1