Developing Multi-Task Learning Methods to Aid in Electronic Healthcare Prediction

Document Type

Article

Abstract

Multi-Task Learning (MTL) is a machine learning paradigm that allows for a neural network model to simultaneously train and predict on multiple related tasks/targets using a shared core network and task-specific branching layers. The related nature of the tasks can be exploited by MTL to regularize the model and its use has been shown to increase performance in all tasks. Imbalanced complexity in tasks leads to MTL models being dominated by a few tasks, which cause conflicting loss function parameters and poor training, resulting in generalized but unspecific models. Medical healthcare records (MHRs) taken on events such as pregnancy often track multiple related targets such maternal morbidity factors (i.e., Sudden Maternal Morbidity, Hemorrhage, Preeclampsia) and long-term health outcomes (i.e., cardiovascular disease, obesity, hypertension, diabetes mellitus, post-partum depression, substance use), which serve as an excellent testbed for the MTL models. This study seeks to develop solutions to task imbalance and interference for MTL systems and demonstrate their efficacy compared to traditional single-task models on a prenatal MHR dataset.

Digital Object Identifier (DOI)

https://doi.org/10.1109/ICHI57859.2023.00077

APA Citation

Tsien, E., Wu, D., & Fede, A. L.-D. (2023). Developing Multi-Task Learning Methods to Aid in Electronic Healthcare Prediction. 2022 IEEE 10th International Conference on Healthcare Informatics (ICHI), 486–487.https://doi.org/10.1109/ICHI57859.2023.00077

Rights

© Copyright 2025 IEEE - All rights reserved, including rights for text and data mining and training of artificial intelligence and similar technologies.

Link to Record

Share

COinS