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

Leveraging multimodal information from biosignals is vital for building acomprehensive representation of people's physical and mental states. However,multimodal biosignals often exhibit substantial distributional shifts betweenpretraining and inference datasets, stemming from changes in task specificationor variations in modality compositions. To achieve effective pretraining in thepresence of potential distributional shifts, we propose a frequency-awaremasked autoencoder ($\texttt{bio}$FAME) that learns to parameterize therepresentation of biosignals in the frequency space. $\texttt{bio}$FAMEincorporates a frequency-aware transformer, which leverages a fixed-sizeFourier-based operator for global token mixing, independent of the length andsampling rate of inputs. To maintain the frequency components within each inputchannel, we further employ a frequency-maintain pretraining strategy thatperforms masked autoencoding in the latent space. The resulting architectureeffectively utilizes multimodal information during pretraining, and can beseamlessly adapted to diverse tasks and modalities at test time, regardless ofinput size and order. We evaluated our approach on a diverse set of transferexperiments on unimodal time series, achieving an average of $\uparrow$5.5%improvement in classification accuracy over the previous state-of-the-art.Furthermore, we demonstrated that our architecture is robust in modalitymismatch scenarios, including unpredicted modality dropout or substitution,proving its practical utility in real-world applications. Code will beavailable soon.

Frequency-Aware Masked Autoencoders for Multimodal Pretraining on Biosignals