Traditionally, inference for graphical dependency structure has assumed stationarity of the estimated distribution. Whilst popular methods for network estimation such as via covariance selection may
