Foundations of Sequence-to-Sequence Modeling for Time Series

TL;DR

This paper provides the first theoretical analysis of sequence-to-sequence models for time series forecasting, comparing them to classical models and guiding practitioners in model selection.

Contribution

It introduces a theoretical framework for sequence-to-sequence time series models and compares their performance to traditional approaches.

Findings

Sequence-to-sequence models have distinct theoretical properties.

Comparison shows advantages and limitations relative to classical models.

Guides practitioners in choosing appropriate modeling techniques.

Abstract

The availability of large amounts of time series data, paired with the performance of deep-learning algorithms on a broad class of problems, has recently led to significant interest in the use of sequence-to-sequence models for time series forecasting. We provide the first theoretical analysis of this time series forecasting framework. We include a comparison of sequence-to-sequence modeling to classical time series models, and as such our theory can serve as a quantitative guide for practitioners choosing between different modeling methodologies.