IMTS is Worth Time $ imes$ Channel Patches: Visual Masked Autoencoders for Irregular Multivariate Time Series Prediction

2025-05-28

AI Summary

Forecasting irregular multivariate time series (IMTS) with non-aligned timestamps and high missingness rates remains a significant challenge. Method: We propose the first vision-inspired masked autoencoder framework specifically designed for IMTS. It introduces a joint time-channel patching scheme with isometric temporal slicing, cross-channel feature completion to mitigate missing-data interference, and a coarse-to-fine two-stage decoder. Additionally, we devise an IMTS-specific self-supervised pretraining strategy. Contribution/Results: Our method achieves substantial improvements over state-of-the-art approaches across multiple IMTS benchmarks, demonstrating strong generalization under low sampling rates and few-shot settings. This work marks the first successful adaptation of vision foundation models to generic time-series modeling, establishing a novel paradigm for high-missingness IMTS forecasting.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Time-Series/Data Streams
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Irregular Multivariate Time Series (IMTS) forecasting is challenging due to the unaligned nature of multi-channel signals and the prevalence of extensive missing data. Existing methods struggle to capture reliable temporal patterns from such data due to significant missing values. While pre-trained foundation models show potential for addressing these challenges, they are typically designed for Regularly Sampled Time Series (RTS). Motivated by the visual Mask AutoEncoder's (MAE) powerful capability for modeling sparse multi-channel information and its success in RTS forecasting, we propose VIMTS, a framework adapting Visual MAE for IMTS forecasting. To mitigate the effect of missing values, VIMTS first processes IMTS along the timeline into feature patches at equal intervals. These patches are then complemented using learned cross-channel dependencies. Then it leverages visual MAE's capability in handling sparse multichannel data for patch reconstruction, followed by a coarse-to-fine technique to generate precise predictions from focused contexts. In addition, we integrate self-supervised learning for improved IMTS modeling by adapting the visual MAE to IMTS data. Extensive experiments demonstrate VIMTS's superior performance and few-shot capability, advancing the application of visual foundation models in more general time series tasks. Our code is available at https://github.com/WHU-HZY/VIMTS.

Problem

• Forecasting Irregular Multivariate Time Series with missing data
• Adapting Visual MAE for sparse multi-channel IMTS prediction
• Improving IMTS modeling via self-supervised learning techniques

Innovation

• Adapts Visual MAE for IMTS forecasting
• Processes IMTS into equal interval patches
• Uses self-supervised learning for IMTS modeling