Modeling Missing Data in Clinical Time Series with RNNs

By Zachary Chase Lipton, David C. Kale, Randall C. Wetzel. 2016

Subfields: Medical & Health AI, Time-Series & Forecasting

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Medical & Health AI

In Time-Series & Forecasting

Citations in the corpus

Deep Reinforcement Learning for Dynamic Treatment Regimes on Medical Registry Data (2017, background): “Some pioneering works [20], [21] are using recurrent neural networks (RNNs) to model the sequence of hospital admission, diagnosis and treatments for broad disease categories with large electronic health record (EHR) datasets.”
Dipole: Diagnosis Prediction in Healthcare via Attention-based Bidirectional Recurrent Neural Networks (2017, background): “Recurrent neural networks (RNNs) can be used for modeling multivariate time series data in healthcare with missing values [6, 18].”
Partially-Supervised Image Captioning (2018, background): “Instead, we focus on the scenario in which the sequences are incomplete or only partially-specified, which occurs in many practical applications ranging from speech recognition [38] to healthcare [39].”
An Improved Multi-Output Gaussian Process RNN with Real-Time Validation for Early Sepsis Detection (2017, methodology): “In addition, (Zhengping et al. (2016)) and (Lipton et al. (2016b)) investigate patterns of informative missingness in physiological ICU time series with RNNs. Finally, our end-to-end technique to discriminatively learn both the MGP and classiﬁer parameters builds off of our prior work (Futoma et…”
Artificial neural networks condensation: A strategy to facilitate adaption of machine learning in medical settings by reducing computational burden (2018, background): “In addition, we add a binary mask input for each variable indicating the time steps that contain a true (vs. imputed) measurement [16].”
Dropout Feature Ranking for Deep Learning Models (2017, methodology): “We following the preprocessing of Lipton et al. [29] work.”
Attend and Diagnose: Clinical Time Series Analysis using Attention Models (2017, methodology): “As a classic tool in time-series analysis, LDS models the linear transition between consecutive states (Liu and Hauskrecht 2013; Liu and Hauskrecht 2016).”
A Survey on the Explainability of Supervised Machine Learning (2020, background): “It is believed that linear models are better interpretable than deep models, where most popular notions of interpretability depend on the comprehensibility of the features (Lipton et al., 2016).”
Multitask learning and benchmarking with clinical time series data (2017, background): “The closest work in spirit to ours is by Collobert and Weston [16], who use a single convolutional network to perform a variety of natural language tasks (part-of-speech tagging, named entity recognition, and language modeling) with diverse sequential structure.”