ShapeCond: Fast Shapelet-Guided Dataset Condensation for Time Series Classification

TL;DR

ShapeCond is a fast, shapelet-guided dataset condensation method that efficiently synthesizes compact time series datasets, preserving local patterns and improving classification accuracy over prior methods.

Contribution

It introduces a novel shapelet-guided optimization framework for time series dataset condensation, significantly improving speed and accuracy.

Findings

29× faster synthesis over CondTSC

Up to 10,000× speedup on long sequences

Outperforms all prior state-of-the-art methods

Abstract

Time series data supports many domains (e.g., finance and climate science), but its rapid growth strains storage and computation. Dataset condensation can alleviate this by synthesizing a compact training set that preserves key information. Yet most condensation methods are image-centric and often fail on time series because they miss time-series-specific temporal structure, especially local discriminative motifs such as shapelets. In this work, we propose ShapeCond, a novel and efficient condensation framework for time series classification that leverages shapelet-based dataset knowledge via a shapelet-guided optimization strategy. Our shapelet-assisted synthesis cost is independent of sequence length: longer series yield larger speedups in synthesis (e.g., 29$\times$ faster over prior state-of-the-art method CondTSC for time-series condensation, and up to 10,000$\times$ over naively using shapelets on the Sleep dataset with 3,000 timesteps). By explicitly preserving critical local patterns, ShapeCond improves downstream accuracy and consistently outperforms all prior state-of-the-art time series dataset condensation methods across extensive experiments. Code is available at https://github.com/lunaaa95/ShapeCond.