Kinguin Affiliate Bestsellers

Time-series data is generated ubiquitously in enormous amounts from sensors and the internet of things. The samples have shape variations that can be incorporated in feature representation and learning. However, large-scale time-series labeling is expensive and requires domain expertise. Therefore, researchers are investigating unsupervised tasks for time series.

Become a football legend with FIFA 22

Machine learning studio. Image credit: Ars Electronica via Flickr, CC BY-NC-ND 2.0

A recent paper on arXiv.org proposes a novel approach of leveraging the large-scale training from a popular computer vision-based dataset to the time-series data in an unsupervised manner. For the first time, the relatedness of ImageNet data and time-series is explored. The approach approaches the problem as the human visual cognitive process by transforming the 1-D time-series data into 2-D images.

The experimental results demonstrate that the proposed method outperforms existing clustering methods without requiring dataset-specific training.

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.

Research paper: Anand, G. and Nayak, R., “Unsupervised Visual Time-Series Representation Learning and Clustering”, 2021. Link: https://arxiv.org/abs/2111.10309




Source link