ReCast: Reliability-aware Codebook Assisted Lightweight Time Series Forecasting

2025-11-14

AI Summary

Real-world time series often exhibit coexisting locality, complexity, and high dynamics, rendering conventional forecasting methods inaccurate and computationally expensive—unsuitable for real-time or resource-constrained deployment. To address this, we propose ReCast, a lightweight forecasting framework. First, it discretizes local patterns via a learnable codebook. Second, it employs a dual-path architecture: a quantized path models regularity, while a residual path captures volatility. Third, it introduces a reliability-aware codebook update mechanism, leveraging distributionally robust optimization (DRO) to dynamically weight and refine the codebook using multi-source reliability signals—enhancing adaptability to non-stationarity and distributional shifts. Extensive experiments demonstrate that ReCast consistently outperforms state-of-the-art methods in prediction accuracy, inference efficiency, and cross-distribution generalization.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Reasoning under Uncertainty: Relational Probabilistic Models

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: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Time series forecasting is crucial for applications in various domains. Conventional methods often rely on global decomposition into trend, seasonal, and residual components, which become ineffective for real-world series dominated by local, complex, and highly dynamic patterns. Moreover, the high model complexity of such approaches limits their applicability in real-time or resource-constrained environments. In this work, we propose a novel extbf{RE}liability-aware extbf{C}odebook- extbf{AS}sisted extbf{T}ime series forecasting framework ( extbf{ReCast}) that enables lightweight and robust prediction by exploiting recurring local shapes. ReCast encodes local patterns into discrete embeddings through patch-wise quantization using a learnable codebook, thereby compactly capturing stable regular structures. To compensate for residual variations not preserved by quantization, ReCast employs a dual-path architecture comprising a quantization path for efficient modeling of regular structures and a residual path for reconstructing irregular fluctuations. A central contribution of ReCast is a reliability-aware codebook update strategy, which incrementally refines the codebook via weighted corrections. These correction weights are derived by fusing multiple reliability factors from complementary perspectives by a distributionally robust optimization (DRO) scheme, ensuring adaptability to non-stationarity and robustness to distribution shifts. Extensive experiments demonstrate that ReCast outperforms state-of-the-art (SOTA) models in accuracy, efficiency, and adaptability to distribution shifts.

Problem

• Addresses ineffectiveness of global decomposition for local dynamic patterns
• Reduces high model complexity for resource-constrained environments
• Enhances robustness against distribution shifts in time series

Innovation

• Encodes local patterns via learnable codebook quantization
• Uses dual-path architecture for regular and irregular modeling
• Employs reliability-aware codebook updates with robust optimization