Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
HypLTSF formalizes multi-scale temporal hierarchies in long-term time series forecasting by explicitly projecting scale-wise representations into the hyperbolic geometry of a Poincaré ball. The framework replaces unconstrained cross-scale interaction with principled geometric regularization, enforcing radial constraints to encode abstraction levels and angular constraints to bind fine-grained dynamics to coarse ancestors. This explicit non-Euclidean inductive bias consistently outperforms existing multi-scale architectures, achieving state-of-the-art forecasting accuracy across benchmark datasets.
Multi-scale temporal hierarchies don't reliably emerge on their own in Euclidean latent spaces, but forcing representations into a Poincaré ball via radial and angular geometric constraints unlocks state-of-the-art long-term forecasting.
Multi-scale modeling has become an effective approach for long-term time series forecasting, capturing temporal patterns that range from fine-grained local dynamics to coarse global trends. Representations across these temporal scales are inherently hierarchical, with coarser scales abstracting and aggregating information from finer ones. While existing approaches readily exchange information across these scales, the hierarchy itself is typically left as an emergent byproduct of such interactions rather than captured as a geometric structure in its own right. In this paper, we introduce HypLTSF, a framework that endows the multi-scale hierarchy with a concrete geometric form by embedding scale-wise representations into the Poincaré ball, whose exponentially expanding volume naturally accommodates hierarchical structures. To align this geometry with the temporal hierarchy, HypLTSF imposes two constraints: (1) a radial constraint that orders embeddings by their level of abstraction, and (2) an angular constraint that groups fine-scale patterns sharing a common coarser-scale ancestor. Extensive experiments on long-term time series forecasting benchmarks show that HypLTSF achieves state-of-the-art performance, suggesting that explicitly modeling the multi-scale hierarchy as a geometric structure is effective for forecasting.