ALER-TI: Aligned Latent Embedding Retrieval for Time Series Imputation

2026-07-08

AI Summary

Existing time series imputation methods often underperform in non-stationary, weakly correlated, or sparse scenarios due to their heavy reliance on local context. This work proposes a retrieval-augmented framework that enhances missing value reconstruction by retrieving similar patterns from historical complete sequences. The key innovation is a Latent Embedding Alignment (LEA) mechanism, which post-processes and aligns corrupted queries with complete candidates in the latent space, effectively mitigating representation mismatch while enabling precomputation and efficient retrieval. Through a lightweight adaptation module, the approach seamlessly integrates with diverse backbone models. Extensive experiments on six real-world datasets demonstrate that the method significantly outperforms strong baselines, achieving superior robustness and imputation accuracy.

Abstract

Deep learning has significantly advanced time series imputation, yet most existing architectures primarily rely on localized temporal context within the corrupted input sequence. This reliance can be limiting in real-world scenarios, where time series often exhibit non-stationary dynamics, weak temporal correlations, and infrequent patterns that are difficult to reconstruct from nearby observations alone. In this paper, we propose ALER-TI, Aligned Latent Embedding Retrieval for Time Series Imputation, a retrieval-augmented framework that explicitly leverages historical patterns to supplement degraded local context for more reliable missing-value reconstruction. The core of ALER-TI is Latent Embedding Alignment (LEA), which mitigates the representation mismatch between corrupted queries and complete historical candidates. By applying post-hoc masking in the latent space, LEA aligns candidates with the query's missingness pattern while allowing historical embeddings to be pre-computed and cached for efficient retrieval. ALER-TI is model-agnostic and can be integrated with various imputation backbones through a lightweight adaptation module. Extensive experiments on six real-world datasets under different missing rates demonstrate that ALER-TI consistently improves strong baseline models and enhances robustness across diverse imputation settings.

Problem

• time series imputation
• missing data
• non-stationary dynamics
• temporal correlation
• historical patterns

Innovation

• retrieval-augmented imputation
• latent embedding alignment
• time series imputation
• non-stationary dynamics
• model-agnostic framework