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

Time series remains one of the most challenging modalities in machinelearning research. The out-of-distribution (OOD) detection and generalizationon time series tend to suffer due to its non-stationary property, i.e., thedistribution changes over time. The dynamic distributions inside time seriespose great challenges to existing algorithms to identify invariantdistributions since they mainly focus on the scenario where the domaininformation is given as prior knowledge. In this paper, we attempt to exploitsubdomains within a whole dataset to counteract issues induced bynon-stationary for generalized representation learning. We propose DIVERSIFY, ageneral framework, for OOD detection and generalization on dynamicdistributions of time series. DIVERSIFY takes an iterative process: it firstobtains the "worst-case" latent distribution scenario via adversarial training,then reduces the gap between these latent distributions. We implement DIVERSIFYvia combining existing OOD detection methods according to either extractedfeatures or outputs of models for detection while we also directly utilizeoutputs for classification. In addition, theoretical insights illustrate thatDIVERSIFY is theoretically supported. Extensive experiments are conducted onseven datasets with different OOD settings across gesture recognition, speechcommands recognition, wearable stress and affect detection, and sensor-basedhuman activity recognition. Qualitative and quantitative results demonstratethat DIVERSIFY learns more generalized features and significantly outperformsother baselines.

DIVERSIFY: A General Framework for Time Series Out-of-distribution Detection and Generalization