Breaking the Context Bottleneck on Long Time Series Forecasting
This addresses the challenge of long-term forecasting in domains like economics and energy, where models often overfit with extended inputs, though it appears incremental as it builds on multiscale modeling methods.
The authors tackled the problem of effectively leveraging long sequences in time-series forecasting by proposing the Logsparse Decomposable Multiscaling (LDM) framework, which decouples patterns at different scales to reduce non-stationarity and improve efficiency, resulting in outperforming all baselines in benchmarks while reducing training time and memory costs.
Long-term time-series forecasting is essential for planning and decision-making in economics, energy, and transportation, where long foresight is required. To obtain such long foresight, models must be both efficient and effective in processing long sequence. Recent advancements have enhanced the efficiency of these models; however, the challenge of effectively leveraging longer sequences persists. This is primarily due to the tendency of these models to overfit when presented with extended inputs, necessitating the use of shorter input lengths to maintain tolerable error margins. In this work, we investigate the multiscale modeling method and propose the Logsparse Decomposable Multiscaling (LDM) framework for the efficient and effective processing of long sequences. We demonstrate that by decoupling patterns at different scales in time series, we can enhance predictability by reducing non-stationarity, improve efficiency through a compact long input representation, and simplify the architecture by providing clear task assignments. Experimental results demonstrate that LDM not only outperforms all baselines in long-term forecasting benchmarks, but also reducing both training time and memory costs.