Network Recovery from Cascade Data: A Debiased Jacobian-Based Machine Learning Approach

Lei Huang|June 5, 2026arXiv

Key Takeaway

You can recover influence networks from cascade data without knowing the underlying diffusion mechanism—CascadeNet uses Jacobian analysis with statistical debiasing to get reliable, model-agnostic network estimates.

Summary

CascadeNet is a machine learning framework that recovers hidden influence networks from cascade data (like disease spread or product adoption) without assuming a specific diffusion model. Using Jacobian-based analysis and debiasing techniques, it provides statistically reliable network estimates that work across different cascade types, validated on COVID-19 transmission data.

training evaluation

Key Terms

jacobian-regularization neyman-orthogonal-debiasing riesz-representer diffusion-model cascade-data