Figure 1
From: Frugal day-ahead forecasting of multiple local electricity loads by aggregating adaptive models
Transfer learning through aggregation of experts in two steps: (a) individual models are trained for E time series for a small E compared to the number of time series, then (b) to forecast any time series k, models \(M_1\), \(M_2\), ..., \(M_E\) are applied and all the resulting forecasts are combined using aggregation of experts.
