TimePerceiver: An Encoder-Decoder Framework for Generalized Time-Series Forecasting

2025-12-27

AI Summary

To address the fragmentation between encoding, decoding, and training in time series forecasting, this paper proposes a unified encoder-decoder framework capable of handling generalized prediction tasks—including extrapolation, interpolation, and missing value imputation. The method introduces three key innovations: (1) a time-aware latent bottleneck encoder that jointly models cross-channel and long-range temporal dependencies; (2) a decoder based on learnable timestamp queries, enabling flexible adaptation to arbitrary input and target time positions; and (3) a multi-task unified training strategy coupled with generic temporal positional embeddings. Evaluated across multiple benchmark datasets, the framework achieves significant improvements over state-of-the-art methods, demonstrating superior predictive accuracy, strong generalization across diverse forecasting tasks, and seamless compatibility with heterogeneous time series structures.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

In machine learning, effective modeling requires a holistic consideration of how to encode inputs, make predictions (i.e., decoding), and train the model. However, in time-series forecasting, prior work has predominantly focused on encoder design, often treating prediction and training as separate or secondary concerns. In this paper, we propose TimePerceiver, a unified encoder-decoder forecasting framework that is tightly aligned with an effective training strategy. To be specific, we first generalize the forecasting task to include diverse temporal prediction objectives such as extrapolation, interpolation, and imputation. Since this generalization requires handling input and target segments that are arbitrarily positioned along the temporal axis, we design a novel encoder-decoder architecture that can flexibly perceive and adapt to these varying positions. For encoding, we introduce a set of latent bottleneck representations that can interact with all input segments to jointly capture temporal and cross-channel dependencies. For decoding, we leverage learnable queries corresponding to target timestamps to effectively retrieve relevant information. Extensive experiments demonstrate that our framework consistently and significantly outperforms prior state-of-the-art baselines across a wide range of benchmark datasets. The code is available at https://github.com/efficient-learning-lab/TimePerceiver.

Problem

• Generalizes forecasting to include extrapolation, interpolation, and imputation tasks
• Handles arbitrarily positioned input and target segments along temporal axis
• Unifies encoder-decoder design with effective training strategy for time-series

Innovation

• Generalized encoder-decoder for diverse temporal prediction tasks
• Latent bottleneck representations capture temporal and cross-channel dependencies
• Learnable queries retrieve information for target timestamps