
The authors propose two interpretable deep learning architectures, TA-RNN and TA-RNN-AE, that leverage time embedding and dual-level attention mechanisms to predict clinical outcomes in electronic health records at the next visit and multiple visits ahead, respectively.


coremsg

attention-based-time-aware-recurrent-neural-network-for-predicting-clinical-outcomes-in-electronic-health-records


Attention-based Time-aware Recurrent Neural Network for Predicting Clinical Outcomes in Electronic Health Records


title_rewrite


SWoTTeD, a novel tensor decomposition method, extracts temporal phenotypes that accurately reconstruct and represent the complex temporal patterns in electronic health records.


temporal-phenotyping-extracting-meaningful-patterns-from-electronic-health-records


Temporal Phenotyping: Extracting Meaningful Patterns from Electronic Health Records
