Time Series Representations for Classification Lie Hidden in Pretrained Vision Transformers

TL;DR

This paper introduces TiViT, a novel approach that converts time series data into images to utilize pretrained Vision Transformers, achieving state-of-the-art classification performance and revealing new insights into representation reuse across domains.

Contribution

The work presents a new framework, TiViT, that leverages pretrained Vision Transformers for time series classification by converting data into images, with theoretical and empirical validation.

Findings

TiViT achieves state-of-the-art results on standard benchmarks.

Intermediate layers with high intrinsic dimension are most effective.

Combining TiViT with TSFM representations improves performance.

Abstract

Time series classification is a fundamental task in healthcare and industry, yet the development of time series foundation models (TSFMs) remains limited by the scarcity of publicly available time series datasets. In this work, we propose Time Vision Transformer (TiViT), a framework that converts time series into images to leverage the representational power of frozen Vision Transformers (ViTs) pretrained on large-scale image datasets. First, we theoretically motivate our approach by analyzing the 2D patching of ViTs for time series, showing that it can increase the number of label-relevant tokens and reduce the sample complexity. Second, we empirically demonstrate that TiViT achieves state-of-the-art performance on standard time series classification benchmarks by utilizing the hidden representations of large OpenCLIP models. We explore the structure of TiViT representations and find that intermediate layers with high intrinsic dimension are the most effective for time series classification. Finally, we assess the alignment between TiViT and TSFM representation spaces and identify a strong complementarity, with further performance gains achieved by combining their features. Our findings reveal a new direction for reusing vision representations in a non-visual domain. Code is available at https://github.com/ExplainableML/TiViT.