Accurate long-term forecasting in complex systems is often compromised by dataset-level distribution shifts, where diverse underlying behavioral modes and evolving system states drive dynamic multivariate time-series. Existing methods predominantly focus on local temporal shifts but fail to explicitly model the global structural challenge where datasets are composites of distinct operational regimes. This paper proposes NEST, a specialized framework designed to model and recompose these evolving structures through a two-phase dense MoE architecture.
NEST first facilitates structural specialization by partitioning the dataset into distinct operational regimes through unsupervised clustering in a principled moment-entropy space. We introduce a regime-oriented router mechanism that generates initial expert weights based on temporal content, subsequently refined through geometric modulation to regime centroids. Crucially, rather than acting as monolithic predictors, individual experts function as specialized kernels capturing regime-specific dynamics by evolving unique variate-attention patterns.
Extensive evaluations on diverse benchmarks, including heterogeneous network traffic and physical phenomena, demonstrate that NEST consistently achieves state-of-the-art performance. Our code and datasets are available at GitHub.
Blogger's Review: NEST effectively tackles dataset-level distribution shifts through its innovative regime-oriented expert mechanism, showcasing significant potential in forecasting complex systems. This approach not only enhances model flexibility but also offers new insights for future research, particularly in handling diverse dynamic data.