Happening @ Michigan https://events.umich.edu/list/rss RSS Feed for Happening @ Michigan Events at the University of Michigan. IOE 899 Seminar Series: Raed Al Kontar (April 13, 2023 3:00pm) https://events.umich.edu/event/107217 107217-21815634@events.umich.edu Event Begins: Thursday, April 13, 2023 3:00pm
Location: Industrial and Operations Engineering Building
Organized By: U-M Industrial & Operations Engineering

Seminar from 3-4 p.m.
Reception to follow in IOE 1727 from 4-5 p.m.

Title: Federated Data Analytics for the Internet of Federated Things (IoFT)

Abstract: A critical change is happening in today's Internet of Things (IoT). The computational power at the edge device is steadily increasing. AI chips are rapidly infiltrating the market. Mobile phones' processing power is becoming comparable to laptops available for everyday use. Tesla's autopilot system has 150 million times more computing power than Apollo 11, and small local computers such as Raspberry Pis have become commonplace in manufacturing systems. This change opens a new paradigm of data analytics within IoT; one that exploits edge compute resources to process more of users' data at the origin of creation. In this talk, I term this future of IoT as the "Internet of Federated Things (IoFT)" and discuss our recent efforts in federated data analytics aimed at bringing this future into reality. Specifically, I will present federated analytics approaches that aim to answer two questions: (1) How to personalize model inference so participants borrow strength from each other yet retain their own individualized models, (2) How to extract what is shared and unique across the distributed datasets? I end the talk by describing our prototyping efforts to generate real-life IoFT data.

Bio: Raed Al Kontar is an Assistant professor in the Industrial and Operations Engineering (IOE) Department at the University of Michigan and an affiliate with the Michigan Institute for Data Science (MIDAS). Raed’s research focuses on collaborative, distributed, and decentralized data science. Raed obtained an undergraduate degree in Civil and environmental engineering and mathematics from the American University of Beirut in 2014. And then a master’s degree in statistics in 2017, and a Ph.D. degree in Industrial & System Engineering in 2018, both from the University of Wisconsin-Madison.

]]>
Workshop / Seminar Wed, 05 Apr 2023 10:45:22 -0400 2023-04-13T15:00:00-04:00 2023-04-13T16:00:00-04:00 Industrial and Operations Engineering Building U-M Industrial & Operations Engineering Workshop / Seminar Raed Al Kontar
Departmental Seminar (899): Mohit Singh (September 7, 2023 3:00pm) https://events.umich.edu/event/110924 110924-21825859@events.umich.edu Event Begins: Thursday, September 7, 2023 3:00pm
Location: Industrial and Operations Engineering Building
Organized By: U-M Industrial & Operations Engineering

Presenter Bio:
Mohit Singh is a Coca-Cola Foundation Professor at the H. Milton Stewart School of Industrial & Systems Engineering (ISyE), Georgia Institute of Technology. His research interests include discrete optimization, approximation algorithms, and convex optimization. His research is focused on optimization problems arising in cloud computing, logistics, network design, and machine learning. Previously, he has worked at Microsoft Research and McGill University and received his PhD in Algorithms, Combinatorics, and Optimization (ACO) program from Tepper School of Business, Carnegie Mellon University in 2008.

Abstract:
Representing data via vectors and matrices and optimizing spectral objectives such as determinants, and traces of naturally associated matrices is a standard paradigm that is utilized in multiple areas including machine learning, statistics, convex geometry, location problems, allocation problems, and network design problems. In this talk, we will look at many of these applications with a focus on the determinant objective. We will then give algorithms for these problems that build on classical matroid intersection algorithms.

]]>
Workshop / Seminar Fri, 25 Aug 2023 12:58:28 -0400 2023-09-07T15:00:00-04:00 2023-09-07T16:00:00-04:00 Industrial and Operations Engineering Building U-M Industrial & Operations Engineering Workshop / Seminar Mohit Singh
899 Seminar Series: Marcia Fampa, Federal University of Rio de Janeiro (September 14, 2023 3:00pm) https://events.umich.edu/event/111912 111912-21827883@events.umich.edu Event Begins: Thursday, September 14, 2023 3:00pm
Location: Industrial and Operations Engineering Building
Organized By: U-M Industrial & Operations Engineering

Abstract:
We develop a branch-and-bound algorithm for the D-optimality problem (D-opt), a central problem in statistical design theory, based on two bounds derived from convex relaxations, the ``natural bound'' and the ``Γ-bound''. We describe a procedure to fix variables based on convex duality. We demonstrate that the Γ-bound for the binary version of D-opt with a particular variable fixing is precisely the ``complementary Γ-bound'' for the so-called ``Data-Fusion problem''. Further, we present theoretical results showing some relations between different bounds. We propose local-search heuristics for D-opt and analyze different strategies to compute search directions and step sizes for each iteration of them. Finally, we present numerical experiments concerning a B&B algorithm based on our results. This is a joint work with Jon Lee and Gabriel Ponte.

Bio:
Marcia Fampa is a Professor at the Federal University of Rio de Janeiro (UFRJ), where she has been since 1997. She is at the Alberto Luiz Coimbra Institute for Graduate Studies and Research in Engineering (COPPE), at UFRJ, where she has supervised more than 32 PhD and master's students. She was a board member of the Brazilian Society of Operations Research (SOBRAPO). Marcia did her undergraduate studies at the Pontifical Catholic University of Rio de Janeiro (PUC/RJ), receiving an engineering degree in 1987. She received her PhD Degree in Systems and Computer Engineering from the Federal University of Rio de Janeiro in 1996. In 1995 and in 2005 she was a visiting researcher at the University of Iowa. She is currently a research visitor at the University of Michigan. Her main research interest is Mixed Integer Nonlinear Programming (MINLP), with a focus on the development of convex relaxations for MINLP problems.

]]>
Workshop / Seminar Fri, 08 Sep 2023 11:01:12 -0400 2023-09-14T15:00:00-04:00 2023-09-14T16:00:00-04:00 Industrial and Operations Engineering Building U-M Industrial & Operations Engineering Workshop / Seminar Marcia Fampa
899 Seminar Series: Daniel Buckland (September 28, 2023 3:00pm) https://events.umich.edu/event/112896 112896-21829721@events.umich.edu Event Begins: Thursday, September 28, 2023 3:00pm
Location: Industrial and Operations Engineering Building
Organized By: U-M Industrial & Operations Engineering

Presenter Bio:

Daniel Buckland MD PhD is an Assistant Professor of Emergency Medicine and Mechanical Engineering at Duke University. He does research in Robotics and Automation as the Director of the Duke Acute Care Technology Lab and serves as the Medical Director of the Laboratory of Transformative Administration, an operational data science group responsible for deploying and maintaining machine learning models in the clinical operations of the Department of Surgery. Dr. Buckland is also the Deputy Chair of the Human System Risk Board of the Office of the Chief Health and Medical Officer via an Intergovernmental Personnel Act agreement with NASA, where he determines the human system risk of spaceflight and how standards, countermeasures, and mission design can mitigate risk.



Abstract:

Autonomy in Safety Critical Systems has long been a goal in healthcare, as well as an underpinning assumption in the progress of human spaceflight from Low Earth Orbit to the exploration of the Moon and Mars. In both contexts the ideal use of autonomy requires the end user to also be the operator or supervisor of the system, whether expertly trained in the operation of the system or not. This talk will present experimental results that explore the development of an autonomous clinical procedure device to make maps of vascular structures in the human body, and the trust untrained operators have in its use. Further applications of these results in the deployment of automated resource allocation in clinical workflows and the use of autonomous medical care in the mitigation of human system risk in spaceflight will also be presented.

]]>
Lecture / Discussion Fri, 22 Sep 2023 14:57:23 -0400 2023-09-28T15:00:00-04:00 2023-09-28T16:00:00-04:00 Industrial and Operations Engineering Building U-M Industrial & Operations Engineering Lecture / Discussion Daniel Buckland
899 Seminar Series: Jasper van der Waa (October 5, 2023 3:00pm) https://events.umich.edu/event/113029 113029-21829940@events.umich.edu Event Begins: Thursday, October 5, 2023 3:00pm
Location: Industrial and Operations Engineering Building
Organized By: U-M Industrial & Operations Engineering

Presenter Bio:

Dr. Jasper van der Waa is a researcher at the Dutch applied research institute TNO on human-AI interaction and collaboration. He heads the Hybrid Intelligence lab on augmenting human capabilities through AI support systems in health care. He leads TNO's research programs on Explainable AI and AI Safety for high-risk applications of AI in health care, the military, and autonomous vehicles. He advises various ministries and companies on how to design and develop AI applications in responsible and human-centric ways. Lastly, he is involved as a senior researcher in various EU Horizon programs focusing on the development of support systems for the maritime domain. During his visit at U-M, he explores possible collaborations between U-M and TNO on the above topics.

Abstract

The collaboration between humans and AI can augment both human and AI capabilities while ensuring the responsible and safe use of AI in safety-critical applications. The design of a successful collaboration requires the combined understanding of human factors and technology embedded in the application context. This talk will present an overview of the work at TNO on human-AI collaboration from this perspective, touching on a wide range of topics such as meaningful control, co-learning, design methodologies, and task allocation. Especially the role of explanations for human-AI collaboration will be discussed. An argument is made on how the field of Explainable AI should broaden its goal from increasing trust to letting explanations serve as the basis of collaboration.

]]>
Lecture / Discussion Mon, 25 Sep 2023 15:14:09 -0400 2023-10-05T15:00:00-04:00 2023-10-05T16:00:00-04:00 Industrial and Operations Engineering Building U-M Industrial & Operations Engineering Lecture / Discussion Jasper van der Waa
899 Seminar Series: Andres Gomez (October 12, 2023 3:00pm) https://events.umich.edu/event/113413 113413-21830972@events.umich.edu Event Begins: Thursday, October 12, 2023 3:00pm
Location: Industrial and Operations Engineering Building
Organized By: U-M Industrial & Operations Engineering

Presenter Bio:

Andrés Gómez received his B.S. in Mathematics and B.S. in Computer Science from the Universidad de los Andes (Colombia). He then obtained his M.S. and Ph.D. in Industrial Engineering and Operations Research from the University of California Berkeley. From 2017 to 2019, Dr. Gómez worked as an Assistant Professor in the Department of Industrial Engineering at the University of Pittsburgh, and since 2019 he is an Assistant Professor in the Department of Industrial and Systems Engineering at the University of Southern California. Dr. Gómez research focuses on developing new theory and tools for challenging optimization problems arising in finance, machine learning and statistics. His research is funded my multiple grants and gifts from the National Science Foundation, the Air Force Office of Scientific Research, Google and Meta.

Abstract

Common statistical techniques fail if the data used to train the model is corrupted by gross errors or outliers. In fact, even the presence of a single outlier may cause estimators to result in arbitrarily large errors. Several robust estimators have been proposed in the statistical literature, which automatically detect and discard outliers before fitting a model using the remaining data. Unfortunately, the resulting training problem is NP-hard and challenging to solve, even with modern optimization techniques. Thus, practitioners typically resort to heuristics, which have inferior statistical properties and may result in low-quality solutions unless stringent assumptions on the data-generation process are made.

]]>
Lecture / Discussion Mon, 02 Oct 2023 19:48:08 -0400 2023-10-12T15:00:00-04:00 2023-10-12T16:00:00-04:00 Industrial and Operations Engineering Building U-M Industrial & Operations Engineering Lecture / Discussion Andres Gomez
899 Seminar Series: Brandon Pitts (November 2, 2023 3:00pm) https://events.umich.edu/event/114347 114347-21832767@events.umich.edu Event Begins: Thursday, November 2, 2023 3:00pm
Location: Industrial and Operations Engineering Building
Organized By: U-M Industrial & Operations Engineering

Presenter Bio:
Dr. Brandon J. Pitts is an Assistant Professor in the School of Industrial Engineering and a Faculty Associate with the Center on Aging and the Life Course (CALC) at Purdue University in West Lafayette, IN. He received a B.S. in Industrial Engineering from Louisiana State University in 2010 and a M.S.E. and Ph.D. in Industrial and Operations Engineering from the University of Michigan (UM) in 2013 and 2016, respectively. Prior to his faculty appointment, he was a Research Fellow at the UM Center for Healthcare Engineering and Patient Safety (CHEPS). Dr. Pitts’ research interests are in the areas of human factors, human-automation/AI interaction, cyber-human-physical systems, interface design, and gerontechnology in complex transportation and work environments, such as driving and aviation. His research has been funded by the National Science Foundation (NSF), Department of Transportation (DOT), Federal Aviation Administration (FAA), National Institute on Disability, Independent Living, and
Rehabilitation Research (NIDILRR), and Ford Motor Company. Dr. Pitts has received several honors and awards, including a 2023 NSF CAREER award and the 1 st place winner ($1M) of the 2022 U.S. DOT Inclusive Design Challenge (IDC) for his team’s EASI RIDER innovation, a life-size autonomous vehicle for individuals with disabilities.



Abstract:
In the coming years, automation and artificial intelligence (AI) will continue to penetrate every area of human life including work, transportation, healthcare, and leisure environments. These advancements promise many benefits, such as improving public safety, extending human abilities, and enabling accessibility. However, as automation and AI become increasingly integrated into our daily activities, several societal challenges and unanswered research questions must be addressed related to the roles/responsibilities of humans interacting with autonomous systems, the impact of automation on human behavior and performance, and the perception of (potential) users of intelligent systems. In this presentation, Dr. Pitts will share insights from a series of research projects aimed to develop and evaluate various autonomous systems in transportation and work environments and for different types of users. Findings from this research program contribute to the broader discussion on how to design effective human automation systems and interfaces for future applications. This work is also helping to develop theories on human perception and performance, shape policies on accessibility, and promote safety in safety-critical environments.

]]>
Lecture / Discussion Mon, 30 Oct 2023 12:49:33 -0400 2023-11-02T15:00:00-04:00 2023-11-02T16:00:00-04:00 Industrial and Operations Engineering Building U-M Industrial & Operations Engineering Lecture / Discussion Brandon Pitts
899 Seminar Series: Alp Muharremoglu (November 16, 2023 3:00pm) https://events.umich.edu/event/115076 115076-21834018@events.umich.edu Event Begins: Thursday, November 16, 2023 3:00pm
Location: Industrial and Operations Engineering Building
Organized By: U-M Industrial & Operations Engineering

Presenter Bio:

Dr. Muharremoglu obtained his BS (1997) in Industrial and Operations Engineering from the University of Michigan and his MS (2000) and PhD (2002) in Operations Research from MIT. After that, he spent 15 years in academia as a faculty member, first as an Assistant Professor at Columbia Business School and later as an Associate Professor at the University of Texas at Dallas’s Naveen Jindal School of Management. In 2017, Dr. Muharremoglu joined the Supply Chain Optimization Technologies group at Amazon, where he is a Senior Principal Scientist.

Abstract:

We develop a method to determine inventory policies in a network, under capacity constraints. A distinguishing feature of the problem is that fulfillment is not determined a priori, and different nodes in the network can satisfy demand in different regions depending on inventory availability, which is important to model in an online retailing environment. We develop a scalable algorithm using Lagrangian decomposition. The model has been launched at Amazon and has led to improvements in important metrics including profitability.

]]>
Lecture / Discussion Thu, 09 Nov 2023 12:32:15 -0500 2023-11-16T15:00:00-05:00 2023-11-16T16:00:00-05:00 Industrial and Operations Engineering Building U-M Industrial & Operations Engineering Lecture / Discussion Alp Muharremoglu
899 Seminar Series: Jamie Gorman, Arizona State University (November 30, 2023 3:00pm) https://events.umich.edu/event/115393 115393-21834616@events.umich.edu Event Begins: Thursday, November 30, 2023 3:00pm
Location: Industrial and Operations Engineering Building
Organized By: U-M Industrial & Operations Engineering

Bio: Dr. Jamie C. Gorman, Ph.D. is a Professor in Human Systems Engineering and Deputy Director of the Center for Human Artificial Intelligence and Robot Teaming (CHART) at Arizona State University and Senior Research Personnel with the NSF Institute for Student-AI Teaming at the University of Colorado. Dr. Gorman’s research focuses on dynamical systems and computational models of team coordination. His research is conducted in complex sociotechnical environments, including medical, space, military, educational, and sports settings, focusing on building generalizable models, metrics, and measurement systems. Dr. Gorman’s research has been funded by DoD, NSF, and industry partners. He is a member of the Human Factors and Ergonomics Society and serves on the editorial boards of Human Factors and the Journal of Experimental Psychology: Applied.

Abstract: Human-autonomy teams operating in dynamic (“perturbed”) environments primarily interact across human and machine elements. However, most measurements are subjective, involving observer ratings and survey responses (e.g., trust; influence; autonomy), and there is a need for theoretically grounded, objective, and metrics of real-time human-machine coordination states. This talk presents generalizable objective measurement systems for measuring the dynamic spread of trust and distrust through influence, quantifying team resilience to automation and autonomy failures, and AI-supported collaborative learning in K -12 education. The practical deployment of measurement frameworks in dashboards and machine learning agents will be discussed.

]]>
Workshop / Seminar Mon, 20 Nov 2023 09:25:48 -0500 2023-11-30T15:00:00-05:00 2023-11-30T16:00:00-05:00 Industrial and Operations Engineering Building U-M Industrial & Operations Engineering Workshop / Seminar Seminar
899 Seminar Series: Qiaomin Xie, U-W Madison ISyE (March 28, 2024 3:00pm) https://events.umich.edu/event/120676 120676-21845119@events.umich.edu Event Begins: Thursday, March 28, 2024 3:00pm
Location: Industrial and Operations Engineering Building
Organized By: U-M Industrial & Operations Engineering

Bio:
Qiaomin Xie is an assistant professor in the Department of Industrial and Systems Engineering at the University of Wisconsin-Madison. Her research interests lie in the fields of reinforcement learning, applied probability, game theory, and stochastic networks, with applications to computer and communication networks. She was previously a visiting assistant professor at the School of Operations Research and Information Engineering at Cornell University (2019-2021). Prior to that, she was a postdoctoral researcher with LIDS at MIT. Qiaomin received her Ph.D. in Electrical and Computing Engineering from the University of Illinois Urbana-Champaign in 2016. She received her B.S. in Electronic Engineering from Tsinghua University. She is a recipient of the NSF CAREER Award, the JPMorgan Faculty Research Award, Google Systems Research Award, and the UIUC CSL Ph.D. Thesis Award.

Abstract:
Many reinforcement/machine learning problems involve loss minimization, min-max optimization, and fixed-point equations, all of which can be cast under the framework of Variational Inequalities (VIs). Stochastic methods like SGD, SEG, and TD/Q Learning are prevalent, and their constant stepsize versions have gained popularity due to effectiveness and robustness. Viewing the iterates of these algorithms as a Markov chain, we study their fine-grained probabilistic behavior. In particular, we establish finite-time geometric convergence of the iterates distribution and relate the ergodicity properties of the Markov chain to the characteristics of the VI, algorithm, and data.

Using techniques of coupling and basic adjoint relationship, we characterize the limit distribution and how its bias depends on the stepsize. For smooth problems, exemplified by TD learning and smooth min-max optimization, the bias is proportional to the stepsize. For nonsmooth problems, exemplified by Q-learning and generalized linear model with nonsmooth link functions (e.g., ReLU), the bias has drastically different behavior and scales with the square root of the stepsize.

This precise probabilistic characterization allows for variance reduction via tail-averaging and bias reduction via Richardson-Romberg extrapolation. The combination of constant stepsize, averaging, and extrapolation provides a favorable balance between fast mixing and low long-run error, and we demonstrate its effectiveness in statistical inference compared to traditional diminishing stepsize schemes.

]]>
Workshop / Seminar Mon, 25 Mar 2024 09:19:58 -0400 2024-03-28T15:00:00-04:00 2024-03-28T16:00:00-04:00 Industrial and Operations Engineering Building U-M Industrial & Operations Engineering Workshop / Seminar Industrial and Operations Engineering Building