Machine Learning for Data-Driven Decisions

Faculty

Jenna Wiens

Jenna Wiens

Associate Professor of Electrical Engineering and Computer Science
2260 Hayward St,
Ann Arbor, MI 48109
University of Michigan
Email: wiensj@umich.edu

I am an Associate Professor of Computer Science and Engineering (CSE). I head the Machine Learning for Data-Driven Decisions (MLD3) research group. My primary research interests lie at the intersection of machine learning and healthcare. I received my Ph.D. in 2014 from MIT.

In 2015, I was named Forbes 30 under 30 in Science and Healthcare; I received an NSF CAREER Award in 2016, and this past year was named to the MIT Tech Review’s list of 35 Innovators Under 35.

PhD Students

Donna Tjandra

Donna Tjandra

Ph.D. Candidate
Email: dotjandr@umich.edu

Erkin Otles

Ph.D. Student
Email: eotles@umich.edu

I’m a third-year Ph.D. student in CSE. My main research interests include developing techniques to leverage electronic health record data and modeling disease progression in the context of semi-competing risks and multiple events. Currently, I am working on a project that adapts existing approaches from survival analysis for competing risks to accommodate semi-competing risks and multiple events. This work is important in a clinical setting because the techniques I develop can be used to identify those at risk of developing specific diseases well before symptom onset.

I am a Medical Scientist Training Program Fellow (MD-PhD student) at the University of Michigan. I have completed the first three years of my medical school training and currently I am in my second year of my PhD work. My research interest lies in creating machine learning and artificial intelligence tools for patients, physicians, and health systems. Most of my work focuses on the development, deployment, and prospective validation of dynamic health outcome prediction models (e.g. early warning systems). I am co-advised by Jenna Wiens (Computer Science and Engineering) and Brian Denton (Industrial and Operations Engineering). I have a professional background in health IT development and hold a Master’s of Engineering from the University of Wisconsin.

Fahad Kamran

Fahad Kamran

Ph.D. Candidate
Email: fhdkmrn@umich.edu

I’m a third-year Ph.D. student in CSE. My main research interests include adapting recent advancements in machine learning to build novel algorithms tailored to survival analysis and causal inference. Currently, I am developing approaches for the accurate estimation of individualized treatment effects to assist clinicians in matching individuals to treatments. In addition, on the more applied side, I explore the use of machine learning and data from wearable sensors to detect, predict, and prevent physiological harm. 

Harry Rubin-Falcone

Harry Rubin-Falcone

Ph.D. Student
Email: hrf@umich

I’m a second-year Ph.D. student in CSE. My research interests include time-series forecasting, machine learning for mental health, and diabetes research. I am currently working on forecasting blood glucose levels in type 1 Diabetes using deep neural networks, with the ultimate goal of facilitating improved blood-glucose management.

Jeeheh Oh

Jeeheh Oh

Ph.D. Candidate
Email: jeeheh@umich.edu

I am a fifth-year Ph.D. student in CSE. My research focuses on the development of novel machine learning algorithms for leveraging clinical time-series data. Clinical data is often limited and using domain knowledge to improve algorithms can benefit performance in these small sample settings. Specifically, I have developed novel architectures that can learn temporal invariances, time- series alignment, as well as to adapt to time-varying tasks. We have used these methods to predict adverse outcomes in clinical settings such as the risk of infection or mortality.

Jiaxuan Wang

Jiaxuan Wang

Ph.D. Candidate
Email: jiaxuan@umich.edu

I’m a fourth-year Ph.D. student in CSE. My main research interests include model interpretability, dataset shift, and non-convex optimization. Currently, I’m developing approaches to tackle covariate shift, which commonly arises in healthcare applications when the demographic composition changes across institutions. Models that can automatically adapt to this kind of shift offer more robustness and ultimately will require less maintenance once deployed.

Jung Min Lee

Ph.D. Student
Email: jminlee@umich.edu

I’m a first-year Ph.D. Student in CSE. My research interest lies in applying machine learning techniques to healthcare problems to create actionable models clincians can use. In the past, I’ve developed a classification model that uses brainwave signals to diagnose insomnia and identify congnitive impairment in patients.

Meera Krishamoorthy

Meera Krishnamoorthy

Ph.D. Student
Email: meerak@umich.edu 

I am a second-year Ph.D. student in CSE. My main research interest is using domain knowledge to inform the creation of novel machine learning methods. My current research project involves developing methods to represent genetic data. In particular, I am exploring the benefits of using machine learning approaches (over existing, non-machine learning approaches) that can leverage context to improve the utility of learned representations.

Sarah Jabbour

Ph.D. Student
Email: sjabbour@umich.edu

I am a first-year Ph.D. student in CSE. My research interests include machine learning and computer vision. Currently, I am focused on developing computer vision models that are robust to unknown biases with applications in medical imaging.

Shengpu Tang

Shengpu Tang

Ph.D. Candidate
Email: tangsp@umich.edu

I’m a first-year Ph.D. student in CSE. My research focuses on developing computational methods for important problems in healthcare, including risk stratification of acute graft-vs-host disease, and dynamic treatment recommendations for acute dyspnea. Currently, I am investigating algorithmic solutions for a clinician-in-the-loop decision support system, by providing more choices to clinicians/patients that help improve patient outcome.

Alumni

Ian Fox

Ian Fox

Ph.D.
Email: ifox@umich.edu

I’m interested in ways to represent and control physiological signals, with a particular emphasis on blood glucose data collected from individuals with diabetes. As diabetes becomes more common throughout the world, such approaches are critical to helping improve outcomes and reduce patient burden. I have developed deep learning-based approaches to multi-step blood glucose forecasting, time-series representation techniques, and novel semi-adversarial approaches to classification. I have also done work using deep reinforcement learning to analyze NBA player tracking data. In my current project, I am working on developing and applying deep reinforcement learning methods to the problem of blood glucose control.  My work has been published in KDD, IJCAI, and ICML.