Time series causal discovery with variable lags

Researchers propose a Tabu-search algorithm for time-series causal discovery that learns edge-specific variable lags rather than assuming fixed lag windows, using decomposable BIC scoring with node-specific effective sample sizes to infer directed acyclic structures from temporal data.