Multitask Learning for Time Series Data with 2D Convolution

TL;DR

This paper explores applying multitask learning to time series classification, proposing a 2D convolutional model that improves expressiveness and outperforms existing methods on benchmark and industrial datasets.

Contribution

It introduces a novel 2D convolution-based model for multitask learning in time series classification, addressing the limitations of 1D convolutional models.

Findings

The 2D convolution model outperforms 1D models on benchmark datasets.

Multitask learning improves generalization in time series classification.

The proposed approach is effective on both public and industrial datasets.

Abstract

Multitask learning (MTL) aims to develop a unified model that can handle a set of closely related tasks simultaneously. By optimizing the model across multiple tasks, MTL generally surpasses its non-MTL counterparts in terms of generalizability. Although MTL has been extensively researched in various domains such as computer vision, natural language processing, and recommendation systems, its application to time series data has received limited attention. In this paper, we investigate the application of MTL to the time series classification (TSC) problem. However, when we integrate the state-of-the-art 1D convolution-based TSC model with MTL, the performance of the TSC model actually deteriorates. By comparing the 1D convolution-based models with the Dynamic Time Warping (DTW) distance function, it appears that the underwhelming results stem from the limited expressive power of the 1D convolutional layers. To overcome this challenge, we propose a novel design for a 2D convolution-based model that enhances the model's expressiveness. Leveraging this advantage, our proposed method outperforms competing approaches on both the UCR Archive and an industrial transaction TSC dataset.