ASIND: Alternating Sparse Identification for Predicting Network Dynamics Without Knowledge

Researchers introduced ASIND, an alternating sparse-identification method for recovering self-dynamics, interaction rules, and network structure without prior model knowledge.

Experiments showed strong performance in identifying network dynamics and making accurate 100-step predictions.

The study also found that very different networks can generate nearly identical trajectories, highlighting weak identifiability from observations alone.

The work could improve interpretable forecasting of complex networked systems while clarifying the limits of reconstructing structure from behavior.