Abstract:Detailed mobile sensing data from phones, watches, and fitness trackers offer an unparalleled opportunity to quantify and act upon previously unmeasurable behavioral changes in order to improve individual health and accelerate responses to emerging diseases. Unlike in natural language processing and computer vision, deep representation learning has yet to broadly impact this domain, in which the vast majority of research and clinical applications still rely on manually defined features and boosted tree models or even forgo predictive modeling altogether due to insufficient accuracy. This is due to unique challenges in the behavioral health domain, including very small datasets (~10^1 participants), which frequently contain missing data, consist of long time series with critical long-range dependencies (length>10^4), and extreme class imbalances (>10^3:1). Here, we introduce a neural architecture for multivariate time series classification designed to address these unique domain challenges. Our proposed behavioral representation learning approach combines novel tasks for self-supervised pretraining and transfer learning to address data scarcity, and captures long-range dependencies across long-history time series through transformer self-attention following convolutional neural network-based dimensionality reduction. We propose an evaluation framework aimed at reflecting expected real-world performance in plausible deployment scenarios. Concretely, we demonstrate (1) performance improvements over baselines of up to 0.15 ROC AUC across five prediction tasks, (2) transfer learning-induced performance improvements of 16% PR AUC in small data scenarios, and (3) the potential of transfer learning in novel disease scenarios through an exploratory case study of zero-shot COVID-19 prediction in an independent data set. Finally, we discuss potential implications for medical surveillance testing.

Multiple task transfer learning with small sample sizes

[Prognostic Model of Small Sample Critical Diseases Based on Transfer Learning].

Multi-Task Learning with Incomplete Data for Healthcare

Multi-Task Learning with Summary Statistics

Exploiting common patterns in diverse cancer types via multi-task learning

Learning Tasks for Multitask Learning: Heterogenous Patient Populations in the ICU

Multi-Task Learning for Alzheimer's Disease Diagnosis and Mini-Mental State Examination Score Prediction

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

Multi-Dataset Multi-Task Learning for COVID-19 Prognosis

Multi-task learning with dynamic re-weighting to achieve fairness in healthcare predictive modeling

Bayesian Meta-Learning for Improving Generalizability of Health Prediction Models With Similar Causal Mechanisms

A Hybrid Instance-based Transfer Learning Method

Clinical Phenotyping Prediction via Auxiliary Task Selection and Adaptive Shared-Space Correction.

Many tasks make light work: Learning to localise medical anomalies from multiple synthetic tasks

Reducing healthcare disparities using multiple multiethnic data distributions with fine-tuning of transfer learning

DBTN: An adaptive neural network for multiple-disease detection via imbalanced medical images distribution

Coupling Deep Imputation with Multitask Learning for Downstream Tasks on Genomics Data

Automated Multi-Task Learning for Joint Disease Prediction on Electronic Health Records

Supervised Transfer Learning at Scale for Medical Imaging

MultiMed: Massively Multimodal and Multitask Medical Understanding

M3H: Multimodal Multitask Machine Learning for Healthcare