RoME for non-stationary HTS

Background: The proposed robust reconciliation method, RoME, is formulated and analyzed under the assumption of stationary time series components. The reconciliation aims to find optimal forecasts by minimizing a general loss function subject to linear aggregation constraints.

Question / Future Work: Extending the proposed framework to hierarchical time series that exhibit non-stationary components is an important direction for future research, as the stationarity assumption is often violated in practical forecasting scenarios. This extension requires further investigation into both the empirical performance and theoretical justification of the RoME method.