Roy Adams

Assistant Professor

Johns Hopkins School of Medicine

I'm looking for a posdoc!

I am currently searching for a postdoc to work on methodological and/or applied projects related to precision medicine for Alzheimer's disease and dementia (full details here). If interested, please email me your CV and the paper you are most proud of.

About Me

I am an Assistant Professor in the Johns Hopkins School of Medicine working on machine learning and statistical methods for electronic health record data as part of the Richman Family Precision Medicine Center of Excellence. Prior to this, I spent three years as a postdoctoral fellow in the Machine Learing and Healthcare lab and the Malone Center for Engineering in Healthcare. I received my PhD in 2018 from University of Massachusetts, Amherst working in the Machine Learning for Data Sciences lab with Prof. Benjamin Marlin.

Research Interests

I am interested in devloping machine learning methods and tools to improve the study of chronic health conditions and improve the safety, efficiency, and efficacy of patient care. Though I am primarily motivated by applications in healthcare, my work addresses core methodological questions in measurement error modeling, causal inference, and model evaluation with broad implications for machine learning and other fields using observational data, such as economics and ecology.

Current Projects

• Measurement error in observational health data

  Data arising from real-world healthcare environments is affected by measurement error and missing data, observed and unobserved confounding, and a variety of societal biases. When reliable measurements are not available, it is critical that we account for systematic measurement error to avoid drawing biased inferences. Unfortunately, observational health data frequently violates the assumptions of classical measurement error analysis and we lack methods to incorporate many measurement error assumptions that commonly occur in healthcare settings. The goal of this project is to develop methods for estimating models from data with measurement error and to derive partial identification bounds when exact model identification is not possible.

• Evaluating machine learning models for use in healthcare environments

  Despite a recent surge in the application of machine learning models to health data, there is still much we do not know about deploying such systems to real healthcare settings. Standard methods for evaluating machine learning models do not tell us how safe a model is to use, the effect a model will have on patient outcomes, or how a model will impact health disparities. The goal of this project is to develop both quantitative and qualitative methods to answer these questions.

Past Projects

• Activity detection from wearable sensors

  Health behaviors such as smoking, eating, drug use, and exercise have traditionally been studied using self-report data; however, self-report has well-known limitations including data sparsity, recall bias, and high burden on study participants. In contrast, wearable sensors allow us to observe patients at high frequency, in non-clinical settings, and without relying on patient self-report. The goal of this project is to develop novel structured predcition models for detecting activities of interest, such as smoking, eating, sleeping, and conversation.

• Targeted Health Messaging

  Behavioral science research has shown that periodical targeted messaging (via email, text message, etc.) can improve the success of smoking secession rates as part of a comprehensive treatment plan. Furthermore, evidence suggests that targeted messaging is more effective if personalized. The goal of this project is to develop and deploy a targeted health messaging system based around current recommender system technologies.

Preprints

* indicates co-first authorship

Katharine E. Henry*, Roy Adams*, Cassandra Parent*, Anirudh Sridharan, Lauren Johnson, David N. Hager, Sara E. Cosgrove, Andrew Markowski, Eili Y. Klein, Edward S. Chen, Maureen Henley, Sheila Miranda, Katrina Houston, Robert C. Linton II, Anushree R. Ahluwalia, Albert W. Wu, and Suchi Saria, “Evaluating Adoption, Impact, and Factors Driving Adoption for TREWS, a Machine Learning-Based Sepsis Alerting System.” preprint

Roy Adams, Suchi Saria, and Michael Rosenblum, “The Impact of Time Series Length and Discretization on Longitudinal Causal Estimation Methods.” preprint

Publications

Noam Finkelstein*, Roy Adams*, Ilya Shpitser, and Suchi Saria, “Partial Identifiability in Discrete Data with Measurement Error,” in Conference on Uncertainty in Artificial Intelligence, 2021. preprint

Adarsh Subbaswamy*, Roy Adams*, and Suchi Saria, “Evaluating Model Robustness to Dataset Shift,” in Conference on Artificial Intelligence and Statistics, 2021. preprint

Roy Adams, Yuelong Ji, Xiaobin Wang, and Suchi Saria, “Learning Models from Data with Measurement Error: Tackling Underreporting,” in International Conference on Machine Learning, 2019. (23% acceptance rate) paper

Roy Adams and Benjamin Marlin, “Learning Time Series Segmentation Models from Temporally Imprecise Labels,” in Conference on Uncertainty in Artificial Intelligence, 2018. (31% acceptance rate) paper

Rummana Bari, Roy Adams, Md. Mahbubur Rahman, Megan Battles Parsons, Eugene Buder, and Santosh Kumar, "rConverse: Moment by moment conversation detection using a mobile respiration sensor," Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, vol. 2, num. 1, pp. 1-27, March 2018. paper

Roy Adams and Benjamin Marlin, “Learning Time Series Detection Models from Temporally Noisy Labels,” in Conference on Artificial Intelligence and Statistics, 2017. (32% acceptance rate) paper code

Thai Nguyen, Roy Adams, Annamalai Natarajan, and Benjamin Marlin, “Parsing Wireless Electrocardiogram Signals with Context Free Grammar Conditional Random Fields,” in IEEE Wireless Health, 2016. paper

Rajani Sadasivam, Erin Borglund, Roy Adams, Benjamin Marlin, and Thomas Houstson. Journal of Medical Internet Research, "Impact of a collective intelligence tailored messaging system on smoking cessation: The PERSPeCT randomized experiment," Journal of Medical Internet Research, vol. 18, num. 11, e285:pp. 1-13, November 2016. paper

Roy Adams, Nazir Saleheen, Edison Thomaz, Abhinav Parate, Santosh Kumar, and Benjamin Marlin, “Hierarchical Span-Based Conditional Random Fields for Labeling and Segmenting Events in Wearable Sensor Data Streams,” in International Conference on Machine Learning, 2016. paper code

Roy Adams, Rajani Sadasivam, Kavitha Balakrishnan, Rebecca Kinney, Thomas Houston, and Benjamin Marlin, “PERSPeCT: Collaborative Filtering for Tailored Health Communications,” in ACM Conference on Recommender Systems, 2014. (24% acceptance rate)

Benjamin Marlin, Roy Adams, Rajani, Sadasivam, and Thomas Houstson, “Towards Collaborative Filtering Recommender Systems for Tailored Health Communications,” in American Medical Informatics Association Annual Symposium, 2013. paper

Refereed Workshop Presentations

Thai Nguyen, Roy Adams, Annamalai Natarajan, and Benjamin Marlin, “Parsing Wireless Electrocardiogram Signals with the CRF-CFG Model,” UAI Workshop: Machine Learning for Health, 2016.

Roy Adams, Edison Thomaz, and Benjamin Marlin, “Hierarchical Nested CRFs for Segmentation and Labeling of Physiological Time Series,” NeurIPS Workshop: Machine Learning for Healthcare, 2015.

Teaching

Previous courses

COMPSCI 590N: Introduction to Numerical Computing with Python

Work Experience

Postdoctoral Researcher, Johns Hopkins University, 2018 to present.
Research Assistant, University of Massachusetts, 2012 to 2018.
Research Intern, Yahoo, Summer 2015
Software Engineering Intern, Hewlett-Packard, Summer 2012
Summer REU, Eco-Informatics Summer Institute, Oregon State University, Summer 2011
Student Programmer, Oregon State University, Summer 2010

Education

PhD in Computer Science, University of Massachusetts, 2018.
MS in Computer Science, University of Massachusetts, 2015.
BS with honors in Computer Science Engineering, University of California - Davis, 2011.