English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

We argue that time series analysis is fundamentally different in nature toeither vision or natural language processing with respect to the forms ofmeaningful self-supervised learning tasks that can be defined. Motivated bythis insight, we introduce a novel approach called \textit{Series2Vec} forself-supervised representation learning. Unlike other self-supervised methodsin time series, which carry the risk of positive sample variants being lesssimilar to the anchor sample than series in the negative set, Series2Vec istrained to predict the similarity between two series in both temporal andspectral domains through a self-supervised task. Series2Vec relies primarily onthe consistency of the unsupervised similarity step, rather than the intrinsicquality of the similarity measurement, without the need for hand-crafted dataaugmentation. To further enforce the network to learn similar representationsfor similar time series, we propose a novel approach that appliesorder-invariant attention to each representation within the batch duringtraining. Our evaluation of Series2Vec on nine large real-world datasets, alongwith the UCR/UEA archive, shows enhanced performance compared to currentstate-of-the-art self-supervised techniques for time series. Additionally, ourextensive experiments show that Series2Vec performs comparably with fullysupervised training and offers high efficiency in datasets with limited-labeleddata. Finally, we show that the fusion of Series2Vec with other representationlearning models leads to enhanced performance for time series classification.Code and models are open-source at\url{https://github.com/Navidfoumani/Series2Vec.}

Series2Vec: Similarity-based Self-supervised Representation Learning for Time Series Classification