AI Insight
Timer-M1 is a new foundation model for multivariate time series forecasting that learns from "primitives"—elementary temporal and relational patterns shared across different domains. The model uses a novel pretraining approach that synthesizes time series data incorporating these primitives and employs gated two-dimensional Transformer blocks to handle relationships between variables. In benchmark testing across three large-scale forecasting datasets, Timer-M1 achieved first-place performance on two benchmarks and second place on a third among recent time series foundation models.
Why it matters
This work addresses a key challenge in time series forecasting: the ability to generalize across different domains without task-specific training. By learning fundamental patterns that exist across diverse applications, the model could enable more accurate predictions in fields ranging from weather forecasting to financial markets and energy demand planning, particularly in scenarios where limited domain-specific training data is available.
Understand the Science
⚠️ Preprint – Noch nicht peer-reviewed
Dieser Artikel wurde noch nicht von unabhängigen Experten begutachtet. Die Ergebnisse sind vorläufig und sollten mit Vorsicht interpretiert werden.
Abstract: We introduce Timer-M1, a pretrained multivariate time series foundation model that learns with primitives for zero-shot forecasting. Across domains, time series share elementary temporal and relational patterns, termed primitives, yet differ in how these primitives manifest and evolve across different contexts. Despite progress in zero-shot and task-general forecasting, existing foundation models may still struggle to generalize to complex real-world scenarios. To this end, we develop a primitive-based data synthesis and pretraining pipeline. The synthesis pipeline generates series with temporal primitives shared across domains and then assembles real and generated series into multivariate samples using relational primitives. Afterwards, samples are organized into episodes by assigning distinct channel roles as target variates, past-only covariates, and known-future covariates, ensuring that the model is optimized on predictable variates using available exogenous information. Technically, Timer-M1 further adapts gated two-dimensional Transformer blocks that dynamically allocate cross-variate attention across layers. Across three large-scale forecasting benchmarks, Timer-M1 ranks first on both FEV and TIME and second on GIFT-Eval among most recent time series foundation models. These results support effective primitive-based pretraining as a route to robust general forecasting technique across domains and task settings.
Source: Timer-M1: A Multivariate Time Series Foundation Model via Learning Primitives