Abstract: Forecasting time-to-event outcomes arises in a variety of healthcare problems such as predicting how long a patient will stay in the hospital or when a disease will relapse. Such problems have been studied for decades in the field of survival analysis. In this talk, I discuss recent progress on making interpretable predictions in survival analysis with the help of a kernel function, which specifies how similar any two patients are. Importantly, how to define similarity depends on the problem at hand. For example, when predicting time until death, which patient attributes matter in defining similarity depends on what diseases the patients have. I explain how to automatically learn such similarity/kernel functions, and how to accurately gauge prediction uncertainty of time-to-event outcomes. I present experimental results on predicting time until death for patients with various illnesses.

Bio: George Chen is an Assistant Professor of Information Systems at Carnegie Mellon University's Heinz College of Information Systems and Public Policy, and an affiliated faculty member of the Machine Learning Department. He primarily works on machine learning for healthcare and for sustainable development, with an emphasis on forecasting problems involving survival analysis as well as time series data. A recurring theme in his work is the use of nonparametric prediction methods that aim to make few assumptions on the underlying data. Since these methods inform interventions that can be costly and affect people’s well-being, ensuring that predictions are reliable is essential. To this end, in addition to developing nonparametric predictors, George also produces theory to understand when and why they work, and identifies forecast evidence to help practitioners make decisions. George received his master’s degree and PhD from MIT in Electrical Engineering and Computer Science, where his PhD thesis won the George M. Sprowls award for best thesis in computer science.

Abstract: Can AI, ML and Data Science help prevent children from getting lead poisoning? Can it reduce infant and maternal mortality? Can it reduce police violence and misconduct? Can it help cities better target limited resources to improve lives of citizens and achieve equity? We're all aware of the potential of ML and AI but turning this potential into tangible and equitable social impact takes cross-disciplinary training, new methods, and dealing with explainability and bias & fairness challenges. In this talk, I'll discuss lessons learned from working on 50+ projects over the past few years with non-profits and governments on high-impact public policy and social challenges in criminal justice, public health, education, economic development, public safety, workforce training, and urban infrastructure. I'll highlight opportunities as well as challenges around explainability and bias/fairness that need to tackled in order to have social and policy impact in a fair and equitable manner.

Bio: Rayid Ghani is a Distinguished Career Professor in the Machine Learning Department and the Heinz College of Information Systems and Public Policy at Carnegie Mellon University. Rayid is a reformed computer scientist and wanna-be social scientist, but mostly just wants to increase the use of large-scale AI/Machine Learning/Data Science in solving large public policy and social challenges in a fair and equitable manner. Among other areas, Rayid works with governments and non-profits in policy areas such as health, criminal justice, education, public safety, economic development, and urban infrastructure. Rayid is also passionate about teaching practical data science and started the Data Science for Social Good Fellowship that trains computer scientists, statisticians, and social scientists from around the world to work on data science problems with social impact.

Before joining Carnegie Mellon University, Rayid was the Founding Director of the Center for Data Science & Public Policy, Research Associate Professor in Computer Science, and a Senior Fellow at the Harris School of Public Policy at the University of Chicago. Previously, Rayid was the Chief Scientist of the Obama 2012 Election Campaign where he focused on data, analytics, and technology to target and influence voters, donors, and volunteers. In his ample free time, Rayid obsesses over everything related to coffee and works with non-profits to help them with their data, analytics and digital efforts and strategy.

Bio: Roni Rosenfeld (B.Sc., mathematics and physics, 1985, Tel-Aviv University; M.Sc. 1991, Ph.D. 1994, computer science, Carnegie Mellon University) is head of the Machine Learning Department and professor of machine learning, language technologies, computer science and computational biology, in the School of Computer Science at Carnegie Mellon University, Pittsburgh, Pennsylvania. He also holds a courtesy appointment at the Heinz School of Public Policy at Carnegie Mellon, and an adjunct appointment at the University of Pittsburgh School of Medicine.

Rosenfeld has been teaching machine learning and statistical language modeling since 1997. He has taught thousands of undergraduate and graduate students, has been a mentor to four post-doctoral students and an advisor to about a dozen Ph.D. students and a score of Masters and undergraduate students.

Professor Rosenfeld's current interests include tracking and forecasting epidemics, using speech and language technologies to aid international development, using machine learning for social good, and advancing data numeracy for all. He has also performed research in statistical language modeling, machine learning, speech recognition and viral evolution. He has published well over 100 scientific articles in academic journals and conferences.

Rosenfeld is a recipient of the Allen Newell Medal for Research Excellence and of the Spira Teaching Excellence Award.

Abstract: Mobility is key to quality of life, equity of opportunity and economic growth in urban environments, and one principal obstacle is poorly timed traffic signals. Real-time traffic signal control presents a challenging multi-agent planning problem, particularly in urban road networks where (unlike simpler arterial settings) there are competing dominant traffic flows that shift through the day. Further complicating matters, urban environments require attention to multi-modal traffic flows (vehicles, pedestrians, bicyclists, buses) that move at different speeds and may be given different priorities. For the past several years, my research group has been evolving an adaptive traffic signal control system to address these challenges, referred to as Surtrac (Scalable Urban TRAffic Control). Combining principles from automated planning and scheduling, multi-agent systems, and traffic theory, Surtrac treats traffic signal control as a decentralized online planning process. In operation, each intersection repeatedly generates and executes (in rolling horizon fashion) signal-timing plans that optimize the movement of currently sensed approaching traffic through the intersection. Each time a new plan is produced (nominally every second) the intersection communicates to its downstream neighbors what traffic it expects to send their way (according to the plan), allowing intersections to construct longer horizon plans and achieve coordinated behavior. Initial deployment of Surtrac in the East End of Pittsburgh has produced significant performance improvements and the technology is now operating in 8 North American cities. More recent work focuses on a broader future vision of smart transportation infrastructure where, as vehicles become more connected and more autonomous, the intersection increasingly becomes the gateway to real-time traffic information and navigation intelligence. Current technology development efforts center on use of direct vehicle- (and pedestrian-) to-infrastructure communication to further enhance mobility, online analysis of traffic flow information for real-time incident detection, and integrated optimization of signal control and route choice decisions. In this talk, I’ll provide an overview of this overall research effort.

Bio: Stephen Smith is a Research Professor in the Robotics Institute at Carnegie Mellon University, where he heads the Intelligent Coordination and Logistics Laboratory. He is also Co-founder and CEO of Rapid Flow Technologies, an intelligent transportations systems (ITS) technology company that is commercializing the Surtrac traffic signal control system. Smith’s research focuses broadly on the theory and practice of next-generation technologies for planning, scheduling, and coordination. He pioneered the development and use of constraint-based search and optimization models for solving planning and scheduling problems, and he has successfully fielded AI-based planning and scheduling systems in several complex application domains. Smith has published over 270 papers on these and related subjects. He recently served as a member of the AAAI Executive Council (2014-2017), is Associate Editor of the Journal of Scheduling, and serves on the editorial boards of Constraints and ACM Transactions on Intelligent Systems and Technology. He was elected AAAI Fellow in 2007.