Unsupervised Visual Time-Series Representation Learning and Clustering

TL;DR

This paper proposes a novel unsupervised learning method for time-series data that leverages data transformation and transfer learning, demonstrating effective clustering performance across diverse applications.

Contribution

Introduces a new unsupervised representation learning approach for time-series using data transformation and transfer learning from large labeled datasets.

Findings

Effective clustering of time-series data demonstrated

Transfer learning improves unsupervised representation quality

Applicable across IoT, sensing, and multimedia domains

Abstract

Time-series data is generated ubiquitously from Internet-of-Things (IoT) infrastructure, connected and wearable devices, remote sensing, autonomous driving research and, audio-video communications, in enormous volumes. This paper investigates the potential of unsupervised representation learning for these time-series. In this paper, we use a novel data transformation along with novel unsupervised learning regime to transfer the learning from other domains to time-series where the former have extensive models heavily trained on very large labelled datasets. We conduct extensive experiments to demonstrate the potential of the proposed approach through time-series clustering.