Network Recovery from Cascade Data: A Debiased Jacobian-Based Machine Learning Approach
Researchers have developed CascadeNet, a novel machine learning framework designed to recover hidden influence networks from cascade data without needing to specify a diffusion model. This approach uses a Jacobian-based method with Neyman-orthogonal debiasing to achieve accurate network inference. CascadeNet demonstrated superior performance in simulations across various data-generating processes and accurately mapped COVID-19 transmission networks in Spain, correlating well with mobility data, unlike existing methods. AI
IMPACT Provides a more robust method for understanding complex diffusion processes, applicable to fields like epidemiology and market analysis.