ReTAMamba: New Model for Irregular Clinical Time Series

Researchers propose ReTAMamba, a novel model for predicting outcomes from irregularly sampled clinical time-series data. The model addresses challenges like missing values and heterogeneous observation patterns by reconstructing data as time-variable token sequences, estimating observation reliability from missingness and elapsed time, and augmenting interval summaries with statistical descriptors. It uses Chronological Weaving to integrate short- and long-term temporal information and a budgeted token router for efficient processing. The approach aims to improve prediction accuracy in clinical settings where data is often irregular.