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

Detailed mobile sensing data from phones, watches, and fitness trackers offeran unparalleled opportunity to quantify and act upon previously unmeasurablebehavioral changes in order to improve individual health and accelerateresponses to emerging diseases. Unlike in natural language processing andcomputer vision, deep representation learning has yet to broadly impact thisdomain, in which the vast majority of research and clinical applications stillrely on manually defined features and boosted tree models or even forgopredictive modeling altogether due to insufficient accuracy. This is due tounique challenges in the behavioral health domain, including very smalldatasets (~10^1 participants), which frequently contain missing data, consistof long time series with critical long-range dependencies (length>10^4), andextreme class imbalances (>10^3:1). Here, we introduce a neural architecturefor multivariate time series classification designed to address these uniquedomain challenges. Our proposed behavioral representation learning approachcombines novel tasks for self-supervised pretraining and transfer learning toaddress data scarcity, and captures long-range dependencies across long-historytime series through transformer self-attention following convolutional neuralnetwork-based dimensionality reduction. We propose an evaluation frameworkaimed at reflecting expected real-world performance in plausible deploymentscenarios. Concretely, we demonstrate (1) performance improvements overbaselines of up to 0.15 ROC AUC across five prediction tasks, (2) transferlearning-induced performance improvements of 16% PR AUC in small datascenarios, and (3) the potential of transfer learning in novel diseasescenarios through an exploratory case study of zero-shot COVID-19 prediction inan independent data set. Finally, we discuss potential implications for medicalsurveillance testing.

Self-supervised Pretraining and Transfer Learning Enable Flu and COVID-19 Predictions in Small Mobile Sensing Datasets