Self-Supervised Learning of Time Series Representation via Diffusion Process and Imputation-Interpolation-Forecasting Mask

TL;DR

This paper introduces TSDE, a novel diffusion-based self-supervised learning method for time series data, which effectively improves various tasks like imputation, forecasting, and anomaly detection by leveraging a new embedding and diffusion process.

Contribution

TSDE is the first diffusion-based SSL approach for time series representation learning, integrating a dual-encoder Transformer and a crossover mechanism for enhanced performance.

Findings

Outperforms existing methods in imputation, forecasting, and anomaly detection

Demonstrates superior embedding quality through visualization and ablation studies

Offers efficient inference speed compared to traditional models

Abstract

Time Series Representation Learning (TSRL) focuses on generating informative representations for various Time Series (TS) modeling tasks. Traditional Self-Supervised Learning (SSL) methods in TSRL fall into four main categories: reconstructive, adversarial, contrastive, and predictive, each with a common challenge of sensitivity to noise and intricate data nuances. Recently, diffusion-based methods have shown advanced generative capabilities. However, they primarily target specific application scenarios like imputation and forecasting, leaving a gap in leveraging diffusion models for generic TSRL. Our work, Time Series Diffusion Embedding (TSDE), bridges this gap as the first diffusion-based SSL TSRL approach. TSDE segments TS data into observed and masked parts using an Imputation-Interpolation-Forecasting (IIF) mask. It applies a trainable embedding function, featuring dual-orthogonal Transformer encoders with a crossover mechanism, to the observed part. We train a reverse diffusion process conditioned on the embeddings, designed to predict noise added to the masked part. Extensive experiments demonstrate TSDE's superiority in imputation, interpolation, forecasting, anomaly detection, classification, and clustering. We also conduct an ablation study, present embedding visualizations, and compare inference speed, further substantiating TSDE's efficiency and validity in learning representations of TS data.