Neural forecasting at scale
This addresses the high memory and computational limitations of current deep ensemble models for forecasting millions of time series in practical scenarios, representing an incremental improvement.
The paper tackles the problem of scaling ensemble-based deep neural networks for multi-step time series forecasting on large datasets by proposing N-BEATS(P), a global parallel variant that reduces training time by half and memory requirements by a factor of 5 while maintaining accuracy.
We study the problem of efficiently scaling ensemble-based deep neural networks for multi-step time series (TS) forecasting on a large set of time series. Current state-of-the-art deep ensemble models have high memory and computational requirements, hampering their use to forecast millions of TS in practical scenarios. We propose N-BEATS(P), a global parallel variant of the N-BEATS model designed to allow simultaneous training of multiple univariate TS forecasting models. Our model addresses the practical limitations of related models, reducing the training time by half and memory requirement by a factor of 5, while keeping the same level of accuracy in all TS forecasting settings. We have performed multiple experiments detailing the various ways to train our model and have obtained results that demonstrate its capacity to generalize in various forecasting conditions and setups.