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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

The present study explores the interpretability of latent spaces produced bytime series foundation models, focusing on their potential for visual analysistasks. Specifically, we evaluate the MOMENT family of models, a set oftransformer-based, pre-trained architectures for multivariate time series taskssuch as: imputation, prediction, classification, and anomaly detection. Weevaluate the capacity of these models on five datasets to capture theunderlying structures in time series data within their latent space projectionand validate whether fine tuning improves the clarity of the resultingembedding spaces. Notable performance improvements in terms of loss reductionwere observed after fine tuning. Visual analysis shows limited improvement inthe interpretability of the embeddings, requiring further work. Results suggestthat, although Time Series Foundation Models such as MOMENT are robust, theirlatent spaces may require additional methodological refinements to beadequately interpreted, such as alternative projection techniques, lossfunctions, or data preprocessing strategies. Despite the limitations of MOMENT,foundation models supose a big reduction in execution time and so a greatadvance for interactive visual analytics.

Decoding Latent Spaces: Assessing the Interpretability of Time Series Foundation Models for Visual Analytics