Happening @ Michigan https://events.umich.edu/list/rss RSS Feed for Happening @ Michigan Events at the University of Michigan. PhD Research Talk: Mohammad Zhalechian (November 1, 2022 3:00pm) https://events.umich.edu/event/100768 100768-21800333@events.umich.edu Event Begins: Tuesday, November 1, 2022 3:00pm
Location: Industrial and Operations Engineering Building
Organized By: U-M Industrial & Operations Engineering

Seminar Abstract:
The rapid growth of information and accessibility to big data provide a unique opportunity to shift toward data-driven decision-making. These real-time paradigms (i) adaptively learn a model that predicts a user-specific outcome for each available decision (prediction) and (ii) harness this model to make data-driven decisions for subsequent users (prescription).

Although there have been tremendous advances in data-driven decision-making, such advances often cannot be applied to operations management problems because of their complexity. This brings forward several challenges and opportunities. In this talk, I discuss two challenges in healthcare and service operations: the need for joint learning and decision-making under limited resources and delayed feedback. I then introduce a data-driven predictive and prescriptive framework with provable performance guarantee to solve a hospital's care unit assignment problem. The effectiveness of this framework is illustrated using hospital system data.

I will end the talk by discussing my broader research agenda on dealing with other practical and societal challenges that arise in developing data-driven decision-making frameworks.


Presenter Bio:
Mohammad Zhalechian is a postdoctoral fellow at the Harvard Kennedy School. His research focuses on data-driven analytics to solve a wide range of problems in healthcare and service operations. He has collaborated closely with hospitals, clinics, and government agencies. Mohammad earned his Ph.D. in Operations Research in Aug 2022 at the University of Michigan, where he was advised by Prof. Mark Van Oyen. He is the recipient of awards, including second place in the 2020 INFORMS Decision Analysis Society Best Paper Award, finalist in the 2020 INFORMS Seth Bonder Scholarship of Health Applications Society, and winner of the 2021 IOE Richard C. Wilson Best Student Paper Award. His research work has also received multiple recognitions in the best paper competitions from MSOM, POMS, and HAS communities.

]]>
Workshop / Seminar Mon, 31 Oct 2022 11:21:32 -0400 2022-11-01T15:00:00-04:00 2022-11-01T16:00:00-04:00 Industrial and Operations Engineering Building U-M Industrial & Operations Engineering Workshop / Seminar Mohammad Zhalechian
PhD Research Talk: Jundi Liu (November 8, 2022 3:00pm) https://events.umich.edu/event/101008 101008-21800665@events.umich.edu Event Begins: Tuesday, November 8, 2022 3:00pm
Location: Industrial and Operations Engineering Building
Organized By: U-M Industrial & Operations Engineering

Over the past decade, autonomous systems have greatly improved how people live and work. However, realizing the full potential of these technologies is only possible if people establish appropriate trust in them. Therefore, prospective systems are expected to sense and respond to users' trust changes and ideally adapt to trust-aware design considerations. I will present my recent work on developing trust-aware customized adaptive systems in vehicle automation using Interactive Reinforcement Learning. We first designed an online driving simulator study to collect human trust dynamics while interacting with vehicle automation. After analyzing trust evolution characteristics, I modeled the trust as a dynamic system using the State Space (SS) model. Then, I proposed an Interactive Reinforcement Learning algorithm to integrate the previously designed trust models into the Inverse Reinforcement Learning (IRL) framework. As a result, the optimal policies recovered by the proposed algorithm can capture the driver preferences of different driving styles from large-scale naturalistic driving data and trust dynamics while interacting with the autonomous systems. Our proposed framework has implications for the design of future human-aware high-fidelity autonomous systems. I will conclude the talk with an overview of how our current work moves toward this future.

Presenter Bio:

Jundi Liu is a Postdoctoral Research Fellow in the Department of Industrial and Operations Engineering at the University of Michigan, working with Prof. Xi Jessie Yang. He graduated from the University of Washington (UW) in September 2022 with a Ph.D. in Industrial and Systems Engineering, working with Prof. Linda Boyle and Prof. Ashis Banerjee on trust-aware customized vehicle automation. Jundi received his B.S. in Computer Science and Engineering from Shanghai Jiao Tong University in 2016 and his M.S. degree in Industrial and Systems Engineering from UW in 2018. Jundi's research interest is to improve human-autonomy interaction through understanding and modeling human trust and to develop trust-aware adaptive systems to support human users in complex decision-making tasks. His research aims to create a holistic framework that enables human-aware high-fidelity autonomous systems.

]]>
Lecture / Discussion Thu, 03 Nov 2022 11:45:16 -0400 2022-11-08T15:00:00-05:00 2022-11-08T16:00:00-05:00 Industrial and Operations Engineering Building U-M Industrial & Operations Engineering Lecture / Discussion Jundi Liu
PhD Researach Talk: Daniel Otero-Leon (November 15, 2022 3:00pm) https://events.umich.edu/event/101202 101202-21800935@events.umich.edu Event Begins: Tuesday, November 15, 2022 3:00pm
Location: Industrial and Operations Engineering Building
Organized By: U-M Industrial & Operations Engineering

Preventing chronic diseases is an essential aspect of medical care. Physicians monitor patients' risk factors and prescribe necessary medication to aim toward better health outcomes, which means that all patients achieve their health potential while accounting for socially and demographic diverse patient populations. Monitoring too frequently may be unnecessary and costly; however, monitoring the patient infrequently means the patient may forgo needed treatment and experience adverse events related to the disease. To accomplish the optimal monitoring policy, we develop models that: (1) Estimate the patient disease progression and (2) Define policies to prevent chronic diseases using sequential decision-making models. We estimate these stochastic models based on longitudinal observational data from electronic health records for a large cohort of patients seen in the national U.S. Veterans Affairs health system. As a result, these models facilitate clinical interventions for cardiovascular diseases, as they help identify high-risk patients. Also, healthcare providers can now rely on additional tools to decide how to manage chronic diseases accurately across demographic subgroups. I will end the talk by discussing how I will build upon this body of work to address healthcare opportunities for improvement.

Presenter Bio:
Daniel F. Otero-Leon is a Ph.D. candidate in the department of Industrial and Operations Engineering (IOE) at the University of Michigan and is co-advised by Dr. Brian Denton and Dr. Mariel Lavieri of IOE. His research interests are generally in operations research and, more specifically, in stochastic models and stochastic dynamic programming with applications to service systems, including health systems and revenue management. His dissertation research is in the area of data-driven models for improving decision-making in the context of cardiovascular disease, with the help of clinical collaborators at the U.S. Department of Veteran Affairs. His work seeks to develop new frameworks that aim for health equity by considering patients' health disparities in disease prevention policies.

]]>
Lecture / Discussion Wed, 09 Nov 2022 10:18:08 -0500 2022-11-15T15:00:00-05:00 2022-11-15T16:00:00-05:00 Industrial and Operations Engineering Building U-M Industrial & Operations Engineering Lecture / Discussion Daniel Otero-Leon
PhD Researach Talk: Rohan Ghuge (November 22, 2022 3:00pm) https://events.umich.edu/event/101458 101458-21801368@events.umich.edu Event Begins: Tuesday, November 22, 2022 3:00pm
Location: Industrial and Operations Engineering Building
Organized By: U-M Industrial & Operations Engineering

Combinatorial optimization captures many natural decision-making problems such as matching, load balancing, assortment optimization, network design, and submodular optimization. In this talk, I will focus on combinatorial problems under uncertainty; specifically when we only have partial knowledge about the input. Solutions to such problems are sequential decision processes that make decisions one by one “adaptively” (depending on prior observations). While such adaptive solutions achieve the best objective, the inherently sequential nature makes them undesirable in many applications. My current research seeks to answer the following: how well can solutions with only a few adaptive rounds approximate fully-adaptive solutions? In this talk, I will formally define the model, and discuss techniques used to answer this question for the stochastic submodular cover problem, which captures problems in domains like sensor placement, medical diagnosis, active learning, and hypothesis testing. I will also state limited adaptivity results that I have obtained for the stochastic score classification and dueling bandits problems. I will conclude the talk with some future work and open problems.

Presenter Bio:
Rohan Ghuge is a Ph.D. candidate in the department of Industrial and Operations Engineering (IOE) at the University of Michigan where he is advised by Dr. Viswanath Nagarajan. His research interests are in optimization under uncertainty, specifically in stochastic combinatorial optimization. His dissertation research explores the role of adaptivity in stochastic combinatorial optimization. He has also worked on designing algorithms for problems arising in domains like assortment optimization, network design and online learning.

]]>
Lecture / Discussion Thu, 17 Nov 2022 16:03:32 -0500 2022-11-22T15:00:00-05:00 2022-11-22T16:00:00-05:00 Industrial and Operations Engineering Building U-M Industrial & Operations Engineering Lecture / Discussion Rohan Ghuge
Ph.D. Research Talk: Haoming Shen (November 29, 2022 3:00pm) https://events.umich.edu/event/101667 101667-21802201@events.umich.edu Event Begins: Tuesday, November 29, 2022 3:00pm
Location: Industrial and Operations Engineering Building
Organized By: U-M Industrial & Operations Engineering

Abstract: In many real-world applications, decision makers usually need to make better decisions without the precise knowledge of the uncertainty. With limited amount of data, distributionally robust chance-constrained optimization (DRC) becomes a powerful tool for decision making because it alleviates the ambiguity in distribution by protecting the optimal solution against a family of candidate distributions, and thus generalizes better when previously unseen samples arise. However, DRC models are usually very hard to solve in general. Therefore, in this talk, I will seek to answer the following two questions: (1) how can we solve DRCs more efficiently, and (2) when are DRCs convex and/or tractable? For DRCs with a covering structure, which arise frequently in facility location, scheduling, production planning, and vehicle routing, we establish their NP-hardness, propose a two-stage reformulation and derive two families of strong valid inequalities. For general DRCs, we uncover a set of sufficient conditions under which DRCs produce a convex feasible region and design efficient algorithms for solving such convex DRCs. I will demonstrate the effectiveness of our proposed solution approaches in multiple real-world applications including the emergency medical facility location problem, optimal power flow problem, and planning of charging stations for battery electric buses.

Bio: Haoming Shen is a Ph.D. Candidate in the Department of Industrial and Operations Engineering at the University of Michigan, where he is advised by Prof. Ruiwei Jiang. His research focuses on data-driven optimization under uncertainty with applications to robotics, power grids, and transportation systems, and has been awarded the honorable mention award in 2022 INFORMS Optimization Society Best Student Paper Competition.

]]>
Presentation Tue, 29 Nov 2022 12:39:32 -0500 2022-11-29T15:00:00-05:00 2022-11-29T16:00:00-05:00 Industrial and Operations Engineering Building U-M Industrial & Operations Engineering Presentation Industrial and Operations Engineering Building
IOE 813 Seminar: Jacqueline Hannan (December 5, 2022 4:30pm) https://events.umich.edu/event/101683 101683-21802222@events.umich.edu Event Begins: Monday, December 5, 2022 4:30pm
Location: Industrial and Operations Engineering Building
Organized By: U-M Industrial & Operations Engineering

Positive pressure ventilation (PPV) provides lifesaving breath support to newborns via a face mask. After birth, approximately 10 million newborns will receive PPV with a face mask every year. As a result, clinicians and hospital workers must be prepared to perform PPV if the infant shows signs of difficulty breathing. PPV requires a careful balance of applied force when holding the face mask against a newborn’s face, as the clinician must provide enough pressure to form a tight seal without delivering too much pressure that could injure the newborn. Currently, there is no standardized tool or quantitative technology to assist in training clinical staff on applying the proper amount of pressure to achieve effective ventilation. In this talk, we discuss current PPV training and administration methods, describe the initial design and testing of a sensor system to monitor applied pressures at key locations on a newborn’s face, and discuss preliminary results from using this sensor system to test the effects of face mask type, ventilation device type, and expertise level on applied pressures.

Jacqueline Hannan is a doctoral student at the University of Michigan in the Industrial and Operations Engineering department. She is studying as a NIOSH trainee and is advised by Prof. Leia Stirling. Prior to attending graduate school, Jacqueline earned her Bachelor of Science in biomedical engineering with a minor in human factors from University at Buffalo. Jacqueline’s research interests lie at the intersection of human factors and healthcare. She is interested in applying concepts of ergonomics and human performance to understand and improve current practices in the medical field.

The seminar series “Providing Better Healthcare through Systems Engineering” is presented by the U-M Center for Healthcare Engineering and Patient Safety (CHEPS): Our mission is to improve the safety and quality of healthcare delivery through a multidisciplinary, systems engineering approach.

For the Zoom link and password, and to be added to the weekly e-mail for the series,
please RSVP, or contact genehkim@umich.edu

]]>
Workshop / Seminar Thu, 01 Dec 2022 11:43:02 -0500 2022-12-05T16:30:00-05:00 2022-12-05T17:30:00-05:00 Industrial and Operations Engineering Building U-M Industrial & Operations Engineering Workshop / Seminar Industrial and Operations Engineering Building
PhD Researach Talk: Seokhyun Chung (December 6, 2022 3:00pm) https://events.umich.edu/event/101702 101702-21802238@events.umich.edu Event Begins: Tuesday, December 6, 2022 3:00pm
Location: Industrial and Operations Engineering Building
Organized By: U-M Industrial & Operations Engineering

Revolutionary advances in Internet of Things (IoT) technologies have realized the emergence of smart & connected systems, where physical units (or devices) that acquire data are connected through a central cloud with intelligence. Increased connectivity across units provides significant opportunities for improving smart analytics through sharing knowledge across units, but, at the same time, reveals critical challenges. First, skyrocketing data scale is beginning to overwhelm computing, storage, and bandwidth resources in the central cloud. Second, personalized analytics is indispensable as units often operate in different environments or have unique features. Third, real-time data collection in smart & connected systems necessitates adapting a smart analytics model to the new data collected.

In this talk, I will present two approaches that address the challenges above, based on a collaborative strategy that distributes learning efforts to the units at the edge while keeping their data stored locally. First, I will discuss a method where physical units collaboratively estimate a multi-output Gaussian process model by exploiting the natural hierarchy of smart & connected systems. Next, I will discuss a two-step approach that (i) finds a well-generalized solution lying in flat minima on the loss surface through distributed computation across units and then (ii) personalizes the learned model to new units without losing old knowledge. Real-world applications for the prediction of lithium-ion battery degradation, condition monitoring signals of turbofan engines, and missing climate data highlight the advantageous features of the proposed models. The talk concludes with some interesting future directions.


Presenter Bio:
Seokhyun Chung is a Ph.D. candidate in the Department of Industrial & Operations Engineering at the University of Michigan, where he is advised by Dr. Raed Al Kontar. His research interests broadly lie in data-driven predictive analytics and decision-making for smart & connected systems. His current research explores collaborative and distributed analytics where Internet of Things-enabled entities (e.g., smartphones, electric vehicles, and wearable devices) exploit their edge computing power to build smart analytics collaboratively. He has been nominated as a finalist in the Quality, Statistics, & Reliability (QSR) Best refereed paper competition at INFORMS 2022 and the Quality Control & Reliability Engineering (QCRE) Best student paper competition at IISE 2021.

]]>
Workshop / Seminar Wed, 30 Nov 2022 10:12:19 -0500 2022-12-06T15:00:00-05:00 2022-12-06T16:00:00-05:00 Industrial and Operations Engineering Building U-M Industrial & Operations Engineering Workshop / Seminar Seokhyun Chung
PhD Research Talk: Xubo Yue (December 13, 2022 3:00pm) https://events.umich.edu/event/101919 101919-21802931@events.umich.edu Event Begins: Tuesday, December 13, 2022 3:00pm
Location: Industrial and Operations Engineering Building
Organized By: U-M Industrial & Operations Engineering

Nowadays, the sheer amount of data collected from edge devices such as mobile phones and self-driving vehicles is beginning to overwhelm traditional centralized data analytics regimes where data from the edge is continuously uploaded to a central server to be processed. Excessive communication traffic from data upload, significant central server storage needs, energy expenditures from centralized learning of big data models, and privacy concerns from sharing raw data are becoming critical challenges in centralized systems. Fortunately, a critical change is happening in today’s Internet of Things (IoT). The processing and computational power of edge devices is becoming increasingly powerful. AI chips are rapidly infiltrating the global market. As such, we now have the opportunity to process more of our data where it is created - i.e., at the edge. This decentralized data analytics paradigm is often coined as federated data analytics (FDA). FDA resolves many of the aforementioned drawbacks. By exploiting edge computations, one can parallelize inference, reduce storage and communication costs, achieve faster alerts and decisions, and protect privacy, amongst many others. Meanwhile, FDA, as an emerging technology, poses significant intellectual challenges. To name a few: (1) most FDA work focuses on deep neural networks and empirical risk minimization (ERM). However, statistical questions such as variable selection, uncertainty quantification, hypothesis testing, and incorporating domain expert knowledge remain unanswered in FDA; (2) edge devices often have local datasets that differ in both size and distribution. Most FDA papers learn a single global model and fail to provide reasonable predictions when heterogeneity exists; (3) IoT systems can raise bias and fairness concerns. Devices with insufficient amounts of data, limited bandwidth, or unreliable internet connection are not favored by conventional training algorithms.
In this talk, I will present my two research papers that address the aforementioned challenges in FDA. First, I will present the federated Gaussian process that provides solutions that go beyond ERM and to correlated settings, which is very common in engineering situations. Interestingly, this model can naturally handle statistical heterogeneity and provide a personalized solution to each edge device. Second, I will present a framework – GIFAIR-FL that imposes group and individual fairness to the FDA setting. The talk concludes with some interesting future directions and my recent work on the collaborative process parameter design and its application in 3D printing


Presenter Bio:
Xubo Yue is a Ph.D. candidate in the Department of Industrial & Operations Engineering at the University of Michigan. His research focuses on federated and distributed data analytics. Currently, he is developing federated data analytics methods that rethink how both prescriptive and predictive analytics are done within IoT-enabled systems, specifically manufacturing and renewable energy. He has received several best paper awards from the Institute for Operations Research and the Management Sciences (INFORMS), the Institute of Industrial and Systems Engineers (IISE), and other renowned organizations.

]]>
Lecture / Discussion Wed, 07 Dec 2022 13:23:40 -0500 2022-12-13T15:00:00-05:00 2022-12-13T16:00:00-05:00 Industrial and Operations Engineering Building U-M Industrial & Operations Engineering Lecture / Discussion Xubo Yue