Category: ECE

Driving in the Snow is a Team Effort for AI Sensors

by Allison Mills, University Marketing and Communications

A major challenge for fully autonomous vehicles is navigating bad weather. Snow especially confounds crucial sensor data that helps a vehicle gauge depth, find obstacles and keep on the correct side of the yellow line, assuming it is visible. Averaging more than 200 inches of snow every winter, Michigan’s Keweenaw Peninsula is the perfect place to push autonomous vehicle tech to its limits.

In two papers presented at SPIE Defense + Commercial Sensing 2021, researchers from Michigan Technological University discuss solutions for snowy driving scenarios that could help bring self-driving options to snowy cities like Chicago, Detroit, Minneapolis and Toronto.

The team includes Nathir Rawashdeh and doctoral student Abu-Alrub (CC) as well as Jeremy Bos and student researchers Akhil Kurup, Derek Chopp and Zach Jeffries (ECE).

Read more about their collaborative mobility research on mtu.edu/news.

This MTU news story was published by Science DailyTechXploreKnowridge Science Report and other research news aggregators.

Dr. Ali Yekkehkhany to Present Talk May 6


Dr. Ali Yekkehkhany, a postdoctoral scholar at the University of California, Berkeley, will present a talk on Thursday, May 6, 2021, at 3:00 p.m.

He will discuss adversarial attacks on the computation of reinforcement learning and risk-aversion in games and online learning.

Dr. Yekkehkhany’s research interests include machine/reinforcement learning, queueing theory, applied probability theory and stochastic processes.

Join the virtual talk here.

Talk Title

Adversarial Reinforcement Learning, Risk-Averse Game Theory and Online Learning with Applications to Autonomous Vehicles and Financial Investments

Talk Abstract

In this talk, we discuss:

  • a) Adversarial attacks on the computation of reinforcement learning: The emergence of cloud, edge, and fog computing has incentivized agents to offload the large-scale computation of reinforcement learning models to distributed servers, giving rise to edge reinforcement learning (RL). By the inherently distributed nature of edge RL, the swift shift to this technology brings a host of new adversarial attack challenges that can be catastrophic in safety-critical applications. A natural malevolent attack could be to contaminate the RL computation such that the contraction property of the Bellman operator is undermined in the value/policy iteration methods. This can result in luring the agent to search among suboptimal policies without improving the true values of policies. We prove that under certain conditions, the attacked value/policy iteration methods converge to the vicinity of the optimal policy with high probability if the number of value/policy evaluation iterations is larger than a threshold that is logarithmic in the inverse of a desired precision.
  • b) Risk-aversion in games and online learning: The fast-growing market of autonomous vehicles, unmanned aerial vehicles, and fleets in general necessitates the design of smart and automatic navigation systems considering the stochastic latency along different paths in a traffic network. To our knowledge, the existing navigation systems including Google Maps, Waze, MapQuest, Scout GPS, Apple Maps, and others are based on minimizing the expected travel time, ignoring the path delay uncertainty. To put the travel time uncertainty into perspective, we model the decision making of risk-averse travelers in a traffic network by an atomic stochastic congestion game and propose three classes of risk-averse equilibria. We show that the Braess paradox may not occur to the extent presented originally and the price of anarchy can be improved, benefiting the society, when players travel according to risk-averse equilibria rather than the Wardrop/Nash equilibrium. Furthermore, we extend the idea of risk-aversion to online learning; in particular, risk-averse explore-then-commit multi-armed-bandits. We use data from the New York Stock Exchange (NYSE) to show that the classical mean-variance and conditional value at risk approaches can come short in addressing risk-aversion for financial investments. We introduce new venues to study risk-aversion by taking the probability distributions into account rather than the summarized statistics of distributions.

Biography

Ali Yekkehkhany is a postdoctoral scholar with the Department of Industrial Engineering and Operations Research, University of California, Berkeley. He received his PhD and MSc degrees in Electrical and Computer Engineering from the University of Illinois, Urbana-Champaign (UIUC) in 2020 and 2017, respectively, and BSc degree in Electrical Engineering from Sharif University of Technology in 2014.

He is the recipient of the “best poster award in recognition of high-quality research, professional poster, and outstanding presentation” in the 15th CSL Student Conference, 2020, and the “Harold L. Olesen award for excellence in undergraduate teaching by graduate students” in the 2019-2020 academic year at UIUC. He was chosen as “teachers ranked as excellent” twice and “teachers ranked as excellent and outstanding” twice at UIUC.

His research interests include machine/reinforcement learning, queueing theory, applied probability theory and stochastic processes.

Dr. Dukka KC, Wichita State, to Present Talk May 5


Dr. Dukka KC, Electrical Engineering and Computer Science, Wichita State University, will present a talk on Wednesday, May 5, 2021, at 3:00 p.m.

Dr. KC will discuss some past and ongoing projects in his lab related to machine learning/deep learning-based approaches for an important problem in Bioinformatics: protein post-translational modification.

Join the virtual talk here.

Talk Title

Bioinformatics as an emerging field of Data Science: Protein post-translation modification prediction using Deep Learning

Talk Abstract

In this talk, I will be presenting about some of the past and ongoing projects in my lab especially related to Machine Learning/Deep Learning based approaches for one of the important problems in Bioinformatics – protein post-translational modification.

Especially, I will focus on our endeavors to get away from manual feature extraction (hand-crafted feature extraction) from protein sequence, use of notion of transfer learning to solve problems where there is scarcity of labeled data in the field, and stacking/ensemble-based approaches.

I will also summarize our future plans for using multi-label, multi-task and multi-modal learning for the problem. I will highlight some of the ongoing preliminary works in disaster resiliency. Finally, I will provide my vision for strengthening data science related research, teaching, and service for MTU’s college of computing.

Biography

Dr. Dukka KC is the Director of Data Science Lab, Director of Data Science Efforts, Director of Disaster Resilience Analytics Center and Associate Professor of Electrical Engineering and Computer Science (EECS) in the Department of EECS at Wichita State University. His current efforts are focused on application of various computing/data science concepts including but not limited to Machine Learning, Deep Learning, HPC, etc. for elucidation of protein sequence, structure, function and evolution relationship among others.

He has received grant funds totaling $4.25M as PIs or Co-PIs, spanning 17 funded grants. He was the PI on the $499K NSF Excellence in Research project focused on developing Deep Learning based approaches for Protein Post-translational modification sites.

He received his B.E. in computer science in 2001, his M.Inf. in 2003 and his Ph.D. in Informatics (Bioinformatics) in 2006 from Kyoto University, Japan. Subsequently he did a postdoc at Georgia Institute of Technology working on refinement algorithms for protein structure prediction. He then moved to UNC-Charlotte and did another postdoc working on functional site predictions in proteins. He was a CRTA Fellow in National Cancer Institute at National Institutes of Health where he was working on intrinsically symmetric domains.

Prior to his arrival at WSU, he was associate professor and graduate program director in the Department of Computational Science and Engineering at North Carolina A&T State University.

Dr. KC has published more than 30 journal and 20 conference papers in the field and is associate editor for two leading journals (BMC Bioinformatics and Frontiers in Bioinformatics) in the field. He also dedicates much of his efforts to K-12 education, STEM workforce development, and increasing diversity in engineering and science.

Sidike Paheding, Applied Computing, Publishes Paper in IEEE Access

A paper co-authored by Sidike Paheding, Applied Computing, has been published in the journal, IEEE Access. “Trends in Deep Learning for Medical Hyperspectral Image Analysis,” was available for early access on March 24, 2021.

The paper discusses the implementation of deep learning for medical hyperspectral imaging.

Co-authors of the paper are Uzair Khan, Colin Elkin, and Vijay Devabhaktuni, all with the Department of Electrical and Computer Engineering, Purdue University Northwest.

Abstract

Deep learning algorithms have seen acute growth of interest in their applications throughout several fields of interest in the last decade, with medical hyperspectral imaging being a particularly promising domain. So far, to the best of our knowledge, there is no review paper that discusses the implementation of deep learning for medical hyperspectral imaging, which is what this work aims to accomplish by examining publications that currently utilize deep learning to perform effective analysis of medical hyperspectral imagery.

This paper discusses deep learning concepts that are relevant and applicable to medical hyperspectral imaging analysis, several of which have been implemented since the boom in deep learning. This will comprise of reviewing the use of deep learning for classification, segmentation, and detection in order to investigate the analysis of medical hyperspectral imaging. Lastly, we discuss the current and future challenges pertaining to this discipline and the possible efforts to overcome such trials.

DOI: 10.1109/ACCESS.2021.3068392

IEEE Access is a multidisciplinary, applications-oriented, all-electronic archival journal that continuously presents the results of original research or development across all of IEEE’s fields of interest. Supported by article processing charges, its hallmarks are a rapid peer review and publication process with open access to all readers.

Emily Zhang Is ME-EM Graduate Seminar Speaker

by Mechanical Engineering – Engineering Mechanics

The next virtual Graduate Seminar Speaker will be held at 4 p.m. tomorrow (Feb. 25) via Zoom.

Lan (Emily) Zhang (ECE) will present “Augmenting Radio Environments for better Wireless Ecosystems.”

Zhang is a member of the Institute of Computing and Cybersystems’s (ICC) Center for Cyber-Physical Systems.

Computing Programs Ranked Among Best in Nation

Several Michigan Tech College of Computing degree programs have been ranked among the best in the nation by Intelligent.com. In addition, the research guide ranked the University number three among all colleges in Michigan.

Intelligent.com looked at nearly 2,300 accredited colleges and universities nationwide making evaluations based on curriculum quality, graduation rate, reputation and post-graduate employment. Programs were evaluated on a scale of 0 to 100 with Michigan Tech making it to the final list for 12 separate degree programs.

The four College of Computing programs and their national ranking as rated by Intelligent.com are:

Additional Michigan Tech degree programs included in the ranking are:

Lan Zhang, ECE, to Present Lecture Jan. 15, 3 pm

Assistant Professor Lan “Emily” Zhang, Electrical and Computer Engineering, will present her lecture, “Augmenting Radio Environments for Better Wireless Ecosystems,” on Friday, January 15, 2021, at 3:00 p.m., via online meeting.

The lecture is hosted by the Michigan Tech Department of Computer Science. Zhang is a member of the Cyber-Physical Systems (CPS) research group of the Institute of Computing and Cybersystems (ICC).

Zhang’s research interest span the fields of cyber-physical systems, distributed machine learning, wireless communications, and cybersecurity. In her talk, she will discuss a series of studies leveraging smart-surfaces, e.g., meta-surfaces or reconfigurable intelligent surfaces (RISs), to augment radio environments for various purposes.

Lecture Abstract

In the last several decades, wireless technologies have become well-established to fight against propagation obstacles. Most conventional efforts are focused on optimizing end devices, such as transmitters and receivers, in order to adapt to the given transmission environment for better communications. However, the recent rapid convergence of the cyber and physical worlds (Cyber-Physical Systems or CPSs) presents unprecedented challenges to the wisdom of conventional design. Given ever-growing service demands, as well as the diverse wireless application scenarios, it is critical to adaptively augment the radio environments in a cost-effective way, while maintaining the aesthetic nature of living environments.

In her talk, Zhang will discuss a series of studies leveraging smart-surfaces–e.g., meta-surfaces or reconfigurable intelligent surfaces (RISs)–to augment radio environments for various purposes. Specifically, she will focus on three promising areas for enhancing the throughput and reliability of wireless communications, mitigating the physical-layer security threats, and facilitating wireless sensing activities. Both model-based and learning-based methods will be used for theoretical and practical analysis.

Biography

Dr. Lan Zhang is an assistant professor in the Department of Electrical and Computer Engineering at Michigan Tech. She received a Ph.D. degree in computer engineering from the University of Florida in 2020, and M.S. and B.Eng. degrees in telecommunication engineering from the University of Electronic Science and Technology of China in 2016 and 2013, respectively.

Zhang has served as a technical program committee member for several respected conferences, such as NeurIPS-SpicyFL 2020 and the 2020 IEEE IFOCOM poster/demo section. She has also served as reviewer for leading journals, such as IEEE Transactions on Communications, IEEE Transactions on Vehicular Technology, IEEE Transactions on Mobile Computing, and IEEE Transactions on Wireless Computing.

Lan Zhang, ECE

ECE Doctoral Defense – Yongyu Wang

by Electrical and Computer Engineering

Computer Engineering doctoral candidate Yongyu Wang will defend at 10 a.m. Tuesday (Dec 1) via Zoom.

The title of his presentation is “High-Performance Spectral Methods for Graph-Based Machine Learning.” Co-advisors are Chee-Wooi Ten (ECE) and Zhuo Feng (ECE).

Chee-wooi Ten is a researcher with the ICC’s Cyber-Physical Systems group.

Hongyu An: Curious About the World and Exploring the Unknown

by Karen S. Johnson, Communications Director, ICC

“A scientist should be a person who is always curious about nature and the world, and who tries to explore the unknown.” –Hongyu An, Assistant Professor, Electrical and Computer Engineering

Hongyu An, Assistant Professor, ECE

Exploring science and technology is always exciting for new Assistant Professor Hongyu An, Electrical and Computer Engineering. He says he is “very pleased to have the chance to mentor the next generation and share my knowledge and experience with undergraduate and graduate students.”

Several things drew Hongyu An to Michigan Tech, including his observation that as an institution Michigan Tech cares about its employees. “The excellent professors, smart students, and the supportive environment are the main reasons I joined Michigan Tech,” he says. “As a new faculty member, I am facing a lot of new challenges. There is great support in my department (ECE) and through the ICC.”

Hongyu is a member of two Institute of Computing and Cybersystems (ICC) research centers: Human-Centered Computing and Scalable Architectures and Systems. He also sees synergies with the Center for Cyber-Physical Systems.

“It is my great pleasure and honor to be a member of the ICC,” Hongyu says. “ I can collaborate with the experts in HCC for exploring the brain and artificial intelligence, and the professors in SAS for hardware and architecture designs. Moreover, the neuromorphic chips I am working on can potentially be applied to Cyber-Physical Systems.”

Hongyu’s primary research area is hardware design for AI and neuromorphic systems. He believes that Artificial Intelligence is probably one of the most challenging research topics in science, noting that recent work in deep learning and artificial neural networks is demonstrating great progress in approaching artificial intelligence. 

“But the traditional computers under von Neumann architecture cannot keep up with the development of neural networks and deep learning,” he cautions. “My research is addressing this challenge by using a new hardware design, from device to architecture levels.”

Hongyu’s teaching interests include VLSI, Circuits, and Electromagnetics. Desribing his teaching philosophy, he notes that making complicated things simple is more challenging than making simple things complicated, and that he strives for the former. This academic year, An is teaching EE 4271 VLSI Design and mentoring ECE master’s student, Sarvani Marthi Sarvani, whose project aims to design a silicon retina through CMOS and Memristors.

Hongyu and his research team are also investigating associative memory learning, a new learning method that aims to create a neuromorphic system that can learn from its surroundings directly. 

“Associative memory is a widespread self-learning method in biological livings, which enables the nervoussystem to remember the relationship between two concurrent events,” Hongyu explains. “Through this learning method, dogs can learn the sound of bells as a sign of food; people can remember a word representing an object.”

“The significance of rebuilding associative memory at a behavioral level not only reveals a way of designing a brain-like, self-learning neuromorphic system, it is also to explore a method of comprehending the learning mechanism of a nervous system,” he adds.

And finally, beyond his work as a professor and scientist Hongyu hopes that he is “a good husband to my wife, a good father to my sons, and a good son to my parents.”

Hongyu completed his Ph.D. in electrical engineering at Virginia Tech, his M.S. in electrical engineering at Missouri University of Science and Technology, and his B.S. in electrical engineering at Shenyang University of Technology.

Recent Publications

An, Hongyu, Mohammad Shah Al-Mamun, Marius K. Orlowski, Lingjia Liu, and Yang Yi. “Robust Deep Reservoir Computing through Reliable Memristor with Improved Heat Dissipation Capability. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems (2020).

An, Hongyu, Qiyuan An, and Yang Yi. “Realizing Behavior Level Associative Memory Learning Through Three-Dimensional Memristor-Based Neuromorphic Circuits. IEEE Transactions on Emerging Topics in Computational Intelligence (2019).

Founded in 2015, the Institute of Computing and Cybersystems (ICC) promotes collaborative, cross-disciplinary research and learning experiences in the areas of computing education, cyber-physical systems, cybersecurity, data sciences, human-centered computing, and scalable architectures and systems, for the benefit of Michigan Technological University and society at large.

The ICC creates and supports an arena in which faculty and students work collaboratively across organizational boundaries in an environment that mirrors contemporary technological innovation. The ICC’s 55 members represent more than 20 academic disciplines at Michigan Tech.