Byte Pair Encoding for Efficient Time Series Forecasting

TL;DR

This paper introduces a pattern-centric tokenization method for time series that adaptively compresses data using motifs, significantly improving forecasting accuracy and efficiency while enabling effective post-hoc optimization.

Contribution

It proposes the first motif-based tokenization scheme for time series, enhancing model performance and efficiency without additional training overhead.

Findings

Improves forecasting performance by 36%.

Increases efficiency by 1990%.

Reduces MSE by up to 44% with conditional decoding.

Abstract

Existing time series tokenization methods predominantly encode a constant number of samples into individual tokens. This inflexible approach can generate excessive tokens for even simple patterns like extended constant values, resulting in substantial computational overhead. Inspired by the success of byte pair encoding, we propose the first pattern-centric tokenization scheme for time series analysis. Based on a discrete vocabulary of frequent motifs, our method merges samples with underlying patterns into tokens, compressing time series adaptively. Exploiting our finite set of motifs and the continuous properties of time series, we further introduce conditional decoding as a lightweight yet powerful post-hoc optimization method, which requires no gradient computation and adds no computational overhead. On recent time series foundation models, our motif-based tokenization improves forecasting performance by 36% and boosts efficiency by 1990% on average. Conditional decoding further reduces MSE by up to 44%. In an extensive analysis, we demonstrate the adaptiveness of our tokenization to diverse temporal patterns, its generalization to unseen data, and its meaningful token representations capturing distinct time series properties, including statistical moments and trends.