Efficient, low regret contextual multi-armed bandit approaches for real time learning including Thompson sampling, UCB, and knowledge gradient descent. Integration of optimization and predictive analytics for determining the time to next measurement, which modality to use, and the optimal control of risk factors to manage chronic disease. Integration of soft voting ensemble classifiers and multiple models Kalman filters for disease state prediction, Real-time (online) contextual multi-armed bandits integrated with optimization of hospital bed type dynamic control decisions for reducing 30-day readmission rates in hospitals. Robustness in system optimization when the system model is uncertain with emphasis on quantile regression forests, sample average approximation, robust optimization and distributionally robust optimization. Health care delivery systems models with prediction and control for inpatient and outpatient. Work has been done on Emergency Department redesign for improved patient flow; Capacity management and planning and scheduling for outpatient care, including integrated services networks; admission control with machine learning to ICUs, stepdown, and regular care units Surgical planning and scheduling for access delay control; Planning and scheduling for Clinical Research Units.
The long temporal and large spatial scales of ecological systems make controlled experimentation difficult and the amassing of informative data challenging and expensive. The resulting sparsity and noise are major impediments to scientific progress in ecology, which therefore depends on efficient use of data. In this context, it has in recent years been recognized that the onetime playthings of theoretical ecologists, mathematical models of ecological processes, are no longer exclusively the stuff of thought experiments, but have great utility in the context of causal inference. Specifically, because they embody scientific questions about ecological processes in sharpest form—making precise, quantitative, testable predictions—the rigorous confrontation of process-based models with data accelerates the development of ecological understanding. This is the central premise of my research program and the common thread of the work that goes on in my laboratory.
Dr. Lee’s research in data science concerns biological questions in systems biology and network medicine by developing algorithms and models through a combination of statistical/machine learning, information theory, and network theory applied to multi-dimensional large-scale data. His projects have covered genomics, transcriptomics, proteomics, and metabolomics from yeast to mouse to human for integrative analysis of regulatory networks on multiple molecular levels, which also incorporates large-scale public databases such as GO for functional annotation, PDB for molecular structures, and PubChem and LINCS for drugs or small compounds. He previously carried out proteomics and metabolomics along with a computational derivation of dynamic protein complexes for IL-3 activation and cell cycle in murine pro-B cells (Lee et al., Cell Reports 2017), for which he developed integrative analytical tools using diverse approaches from machine learning and network theory. His ongoing interests in methodology include machine/deep learning and topological Kolmogorov-Sinai entropy-based network theory, which are applied to (1) multi-level dynamic regulatory networks in immune response, cell cycle, and cancer metabolism and (2) mass spectrometry-based omics data analysis.
S. Sriram, PhD, is Associate Professor of Marketing in the University of Michigan Ross School of Business, Ann Arbor.
Prof. Sriram’s research interests are in the areas of brand and product portfolio management, multi-sided platforms, healthcare policy, and online education. His research uses state of the art econometric methods to answer important managerial and policy-relevant questions. He has studied topics such as measuring and tracking brand equity and optimal allocation of resources to maintain long-term brand profitability, cannibalization, consumer adoption of technology products, and strategies for multi-sided platforms. Substantively, his research has spanned several industries including consumer packaged goods, technology products and services, retailing, news media, the interface of healthcare and marketing, and MOOCs.
My interests are in the areas of labor economics, program evaluation, and the economics of education. Currently my research focuses on college student debt accumulation and the subsequent risk of default, the effect of tuition subsidies on college attendance, the influence of family wealth on college attendance and completion, the effect of financial aid packages on college attendance, completion and subsequent labor market earnings, the influence of education on job displacement and subsequent earnings, the impact of unemployment insurance rules on unemployment durations and re-employment wages, and the determinants and consequences of repeat use of the unemployment insurance system.
Sriram Chandrasekaran, PhD, is Assistant Professor of Biomedical Engineering in the College of Engineering at the University of Michigan, Ann Arbor.
Dr. Chandrasekaran’s Systems Biology lab develops computer models of biological processes to understand them holistically. Sriram is interested in deciphering how thousands of proteins work together at the microscopic level to orchestrate complex processes like embryonic development or cognition, and how this complex network breaks down in diseases like cancer. Systems biology software and algorithms developed by his lab are highlighted below and are available at http://www.sriramlab.org/software/.
– INDIGO (INferring Drug Interactions using chemoGenomics and Orthology) algorithm predicts how antibiotics prescribed in combinations will inhibit bacterial growth. INDIGO leverages genomics and drug-interaction data in the model organism – E. coli, to facilitate the discovery of effective combination therapies in less-studied pathogens, such as M. tuberculosis. (Ref: Chandrasekaran et al. Molecular Systems Biology 2016)
– GEMINI (Gene Expression and Metabolism Integrated for Network Inference) is a network curation tool. It allows rapid assessment of regulatory interactions predicted by high-throughput approaches by integrating them with a metabolic network (Ref: Chandrasekaran and Price, PloS Computational Biology 2013)
– ASTRIX (Analyzing Subsets of Transcriptional Regulators Influencing eXpression) uses gene expression data to identify regulatory interactions between transcription factors and their target genes. (Ref: Chandrasekaran et al. PNAS 2011)
– PROM (Probabilistic Regulation of Metabolism) enables the quantitative integration of regulatory and metabolic networks to build genome-scale integrated metabolic–regulatory models (Ref: Chandrasekaran and Price, PNAS 2010)
Jon Zelner, PhD, is Assistant Professor in the department of Epidemiology in the University of Michigan School of Public Health. Dr. Zelner holds a second appointment in the Center for Social Epidemiology and Population Health.
Dr. Zelner’s research is focused on using spatial analysis, social network analyisis and dynamic modeling to prevent infectious diseases, with a focus on tuberculosis and diarrheal disease. Jon is also interested in understanding how the social and biological causes of illness interact to generate observable patterns of disease in space and in social networks, across outcomes ranging from infection to mental illness.
The goal of my research is to leverage network analysis techniques to uncover how the brain mediates sex hormone influences on gendered behavior across the lifespan. Specifically, my data science research concerns the creation and application of person-specific connectivity analyses, such as unified structural equation models, to time series data; these are intensive longitudinal data, including functional neuroimages, daily diaries, and observations. I then use these data science methods to investigate the links between androgens (e.g., testosterone) and estradiol at key developmental periods, such as puberty, and behaviors that typically show sex differences, including aspects of cognition and psychopathology.
People’s behavior is often contingent on what other people are doing or have done. In dating and job markets, for example, each person’s choices limit what opportunities are available to others. A classic problem in sociology is explaining the relationship between individuals’ actions and larger-scale social patterns. My strategy is to use computer models of how people’s choices co-evolve with aspects of their environment—known as agent-based models (ABMs)—to determine what behavioral or demographic features are important for understanding social processes. I then use statistical models to assess to what degree these features exist in the real world. Substantively, most of my work examines the drivers of neighborhood segregation. More recently, I embarked on a study of how mate choice strategies shape (and are shaped by) dating, marriage, and affair markets.
With Fred Feinberg (UM Marketing and Statistics), I am also exploring how new data sources can be combined with choice models. The vast amounts of activity data from sources such as cell phones and the Internet make it possible to study human behavior with an unparalleled richness of detail. Such “big data” are interesting in large part because they are behavioral data that allow us to observe how people explore their environment, engage in novel or habitual behaviors, interact with others, and learn from past experiences. In ongoing work, we show how decision processes regarding mate choice can be extracted from online dating activity data.
Andrzej Galecki, MD, PhD, is Research Professor in the department of Biostatistics, School of Public Health, and Research Professor in the Institute of Gerontology at the University of Michigan, Ann Arbor.