Alleviating RoME compute burden

Background: The robustness of M-estimation based forecast reconciliation for Hierarchical Time Series (HTS) is evaluated through various simulation scenarios and real-data applications. The proposed method, RoME, generalizes reconciliation procedures by employing a general loss function to mitigate the impact of irregular or “bad” base forecasts.

Question / Future Work: Incorporating penalized estimation methods, such as those used in high-dimensional covariance estimation, might help alleviate the substantial computational burden incurred by the RoME framework when applied to cross-sectional hierarchical time series.