Summer School on AI & Smart Health 2020 - Lecture Details

NSF Track 2 CRESH: Summer School on AI & Smart Health 2020

Monday July 20 - Tuesday 21, 2020. Lecture Details Below

Lecture Details

[Day 1, Topic 1] Introduction to Machine Learning

This module begins by introducing what can be achieved today with Machine Learning. It will then overview the basics of supervised learning. In particular, it will discuss the notions of features, linear predictors, loss minimization, and stochastic gradient descent. It will do so within the context of a typical machine learning workflow that entails the collection of data, the learning, and the validation prior to a practical deployment of model predictors.

Speaker(s)

Dr. Gianfranco Doretto

Gianfranco Doretto joined West Virginia University in 2010, where he is now an Associate Professor. His research interests span several areas of computer vision and machine learning, with a current focus of using deep learning techniques for learning representations that supports applications such as video analysis, biometrics, robotics, and smart health. Before joining WVU, Dr. Doretto was a lead scientist at GE Global Research. He holds a D.Eng. degree in Electronics Engineering from the University of Padua, Italy, and a M.S. and a Ph.D. in Computer Science, both from UCLA.

[Day 1, Topic 2] Social Network Analysis

Social media has become an indispensable communication tool giving rise to global online social networks. Analyzing social media data helps study the properties and behaviors of individuals, communities, and organizations and advance our understanding of social and cultural dynamics through the lens of modern information and communication technologies (ICTs). The knowledge thus extracted helps in harnessing business intelligence and guides decision making. The course will introduce basic social science concepts, theories, and principles relevant to guide model development, data analysis, and inference making. Data collection from social media sites using crawlers and APIs will be discussed. Interfacing the collected data with visualization and analytical tools will be taught to enable better understanding of data and extracting actionable knowledge. Specifically, social network analysis (SNA) techniques will be taught to meaningfully study the data to identify trends, leaders, and groups. These skills can be used to analyze social media data about products, services, campaigns, markets, events, customers, and employees; segment audience by geography or demographics, influencers, recommenders or detractors; and measure social media activities. Case-studies will be discussed to demonstrate the impact even the most basic analyses could afford.

Lesson plan for SNA

Social Network analysis course

Intro/Motivation (30 mins)
Social Network Analysis + Graph theory (lecture) - centrality, clustering (60 mins)
Break (10 mins)
Case Study (Gephi visualization) (40 mins)
- Twitter Data Collection (10 mins)
- Twitter Gephi Visualization (20 mins)
- YouTube Trending videos visualization (10 mins)

Speaker(s)

Dr. Nitin Agarwal

Dr. Nitin Agarwal is Maulden-Entergy Endowed Chair and Distinguished Professor of Information Science at University of Arkansas - Little Rock. He is the founding director of the Collaboratorium for Social Media and Online Behavioral Studies (COSMOS). His research contributions lie at the intersection of social computing, behavior-cultural modeling, collective action, social cyber forensics, artificial intelligence, data mining, and machine learning. His research aims to push the boundaries of our understanding of digital and cyber social behaviors that emerge and evolve constantly in the modern information and communication platforms. From Saudi Arabian women's right to drive cyber campaigns to Autism awareness campaigns to ISIS' and anti-West/anti-NATO disinformation campaigns, at COSMOS, he is directing several projects with over $15 million in funding that have made foundational and applicational contributions to social and computational sciences, particularly in understanding coordinated cyber campaigns. He has published 10 books and over 200 articles in top-tier peer-reviewed forums including the NATO’s Defense StratCom Journal, with several best paper awards and nominations. His most recent book explores the deviant behaviors on the Internet and is published by Springer in their series on cybersecurity. His work has been covered by local, national, and international media including Bloomberg, US News, KUAR, Arkansas Business, Arkansas Times, Arkansas Democrat Gazette, and many others. Over the last several years, Dr. Agarwal has spoken at various public and professional, national and international forums such as the NATO’s StratCom COE (Riga, Latvia), DARPA, US Department of State, US Naval Space and Warfare (SPAWAR), US Pentagon’s Strategic Multilevel Assessment groups, US National Academies of Sciences Engineering and Medicine, US Office of the Director of National Intelligence, Facebook Asia Pacific HQ, Twitter Asia Pacific HQ, US Embassy in Singapore, Singapore Ministry of Communication and Information, NATO Senior Leadership meetings, USIP, among others. Visit https://profiles.ualr.edu/na10 for more details.

Dr. Muhammad Nihal Hussain

Dr. Muhammad Nihal Hussain is a postdoctoral fellow at COSMOS lab at University of Arkansas - Little Rock. His research interest includes multi-platform disinformation campaigns, cross-media information diffusion, crowd manipulation strategies on content-centric social media platforms (Blogs and YouTube). Dr. Hussain has studied several online campaigns ranging from analyzing content-centric platforms to extract opinions and adversarial narrative around NATO's various events and exercises to understanding role of individual social media platforms in a complex multi-platform campaign. He has published 2 book chapters, over 30 peer-reviewed articles, and delivered tutorials and invited talks on blogosphere analysis at SBP-BRiMS, NATO TIDE Sprint and "Social media course" organized by NATO STRATCOM COE in Riga, Latvia. Visit https://cosmos.ualr.edu/about/nihal/ for more details.

[Day 1, Topic 3] Introduction to Deep Learning Architectures

This lecture provides a broad introduction to the basic concept of the classical neural networks (NN) and its current evolution to deep learning (DL) technology. The primary goal of this lecture is to introduce the well-known deep learning architectures and their applications. This lecture will describe the history of neural networks and its progress to current deep learning technology. It covers several deep learning architectures such the classical multi-layer feed forward neural networks, multi-layer auto-encoders, and examples of popular convolutional neural network-based architectures such as AlexNet, VGGNet, GooGleNet (inception modules), and ResNet. Advanced architectures such as Siamese deep networks, coupled neural networks will also be covered.

Speaker(s)

Nasser M. Nasrabadi

Nasser M. Nasrabadi is a professor in the Lane Computer Science and Electrical Engineering Department at West Virginia University. He was senior research scientist (ST) at US Army Research Laboratory (ARL). He is actively engaged in research in deep learning, biometrics, computer vision and image processing. He has published over 300 papers in journals and conference proceedings. He has been an associate editor for the IEEE Transactions on Image Processing, IEEE Transactions on Circuits and Systems for Video Technology and IEEE Transactions for Neural Networks. He is a Fellow of IEEE and SPIE.

[Day 2, Topic 4] AI in Smart Health: Cardiovascular Imaging and Data Analysis

Cardiovascular medicine generates a plethora of biomedical, clinical and imaging data as part of patient care delivery. Often these data are stored in diverse data repositories which are not readily utilizable for cardiovascular research due to challenges in automated abstraction and manual curation technical competency. Moreover, clinical decisions for patient care from data are limited to visual and quantitative assessment of cardiac structure and function. However, it is becoming evident that the best way to make decisions on the basis of data is through the application of machine learning and artificial intelligence techniques.

Application of traditional machine learning techniques in cardiology, however, are challenging due to the inherent difficulties in understanding large, high-dimensional and, often, noisy unlabeled data. Specifically, traditional machine learning techniques, despite their effectiveness, still have some drawbacks which might introduce unwanted biases (e.g., specific metrics). Often the success of the machine learning model depends on how well it fits the underlying data. In this lecture, we will introduce a new approach of understanding patterns in the data that are associated with its shape, topological data analysis (TDA). TDA provides a general framework for analyzing data to identify groups or patterns that provide more information about structure in the logical relationships among phenotypic, demographic, and disease variables which can then be combined with traditional machine learning approaches. This lecture will give students a theoretical/fundamental and practical experience on TDA in particular what it is, how it works, and how it can work with machine learning methods to get better results than are possible by either technique individually.

Course objectives:

Teaching participant the basic underlying theory behind topological data analysis.
Teaching participant how to identify a way that highlights high value segments of the data through TDA using a demo.

Speaker(s)

Partho Sengupta, MD, MBBS, FACC, FASE

Chief of the Division of Cardiology, Director of Cardiovascular Imaging, and Chair of Cardiovascular Innovation at WVU Heart and Vascular Institute; Professor of Medicine in the WVU School of Medicine.

Prior to joining WVU Medicine, Dr. Sengupta was the director of interventional echocardiography, cardiac ultrasound research, and core lab at Mount Sinai’s Zena and Michael A. Weiner Cardiovascular Institute and the Marie-Josée and Henry R. Kravis Center for Cardiovascular Health. Dr. Sengupta received his medical degree from Government Medical College in India, where he also completed a residency. He then completed a cardiology fellowship from All India Institute of Medical Sciences and a cardiology fellowship at both Mayo Clinic in Minnesota and Arizona. Dr. Sengupta is the associate editor of the Journal of American College of Cardiology: Cardiovascular Imaging, section editor for the Journal of American College of Cardiology, on the editorial board of eight cardiology journals, and has more than 200 publications and text book chapters. He is the current chair of Innovation Task Force at the American Society of Echocardiography. He has been a TedMed speaker and has been recognized as a Top 25 professor of ultrasound medicine. Dr. Sengupta is board certified in internal medicine, cardiovascular medicine, and echocardiography.

Naveena Yanamala, Ph.D.

Naveena Yanamala, Ph.D., is the Principal Data Scientist in the Heart & Vascular Institute at WVU Medicine, where she has been since April 2020. In addition, she currently holds an adjunct professor appointment in the Institute for Software Research, School of Computer Science at the Carnegie Mellon University and in the Center for Computational Natural Sciences and Biology, International Institute for Information Technology (IIIT). From 2012 to 2020, she was a Research Biologist (Service Fellow) in the Health Effects Laboratory Division at National Institute for Occupational Safety and Health, Morgantown. Dr. Yanamala also served as a steering committee member of center for occupational robotics research (CORR) and as liaison for machine learning and artificial intelligence in a NIOSH wide emerging technologies interest group (ETIG), while maintaining an active collaboration with various institutions, including University of Pittsburgh, West Virginia University Health Sciences Center, Carnegie Mellon University and several different research teams across NIOSH with a major focus on applied AI. Dr. Yanamala currently serves on the Editorial Board of International Journal of Nanomaterials, Nanotechnology and Nanomedicine.

Dr. Yanamala has 13+ years of experience in conducting effective interdisciplinary research at the intersection of biology, health, and computation. As an applied ML/AI researcher, her research efforts have focused on issues as varied as (a) identifying factors that mitigate toxicity to enable safe-design of novel engineered materials, (b) developing biomarkers of exposure/disease, to classify or predict disease outcomes in environmental/occupational settings, (c) automation of medical diagnostic processes and healthcare forecasting (e.g., pathology, radiology, scheduling) and (d) scalable solutions for intelligent automation in public health surveillance. Most of her research has been at the heart of such endeavors ranging from providing improved insights through to automation, combining informatics, data analytics, computational biology, toxicology, and medical sciences. This has led to more than 60 journal articles, 4 review articles and 3 conference papers in international and national scientific journals.

She received her Ph.D. in Integrative Systems Biology in 2009 from University of Pittsburgh and her MS in Information Technology with specialization in Bioinformatics, from International Institute of Information Technology, Hyderabad, India along with an internship in ML from Carnegie Mellon University in 2004. She possesses a Bachelor of Science degree from S. V. University in Computer Science.

[Day 2, Topic 5] AI, Privacy & Fairness in Healthcare

The remarkable development of AI in healthcare domain presents obvious privacy and fairness issues, when data analytics models are built on users’ personal and highly sensitive data, e.g., clinical records, user profiles, and biomedical images. In this talk, we start with the general problem of privacy, medical ethics, and their significance in digital health. We then concentrate on research of differential privacy preserving machine learning. Differential privacy ensures that the adversary cannot infer any information about any particular record with high confidence (controlled by a privacy budget) from the released learning models. We introduce the concept of differential privacy and present several mechanisms, including Laplace mechanism, exponential mechanism, input perturbation, and functional perturbation, that have been developed to enforce differential privacy in machine learning models, and discuss how to apply and adapt those mechanism to preserve differential privacy in deep learning models. Next we focus on fairness aware machine learning as decision models can cause discrimination against individuals and groups beyond damages to privacy. We introduce a unified causal modeling and counterfactual inference based framework to capture and measure discrimination accurately and present challenges and several efficient algorithms to discover and remove various types of discrimination such as direct/indirect and individual/group/system discrimination from historical training datasets as well as predictive models while preserving data utility. Some recent work of fairness aware generative adversarial networks and fair recommendation will also be covered. We conclude the talk with challenges of building trustworthy AI systems in digital health.

Speaker(s)

Dr. Xintao Wu

Dr. Xintao Wu is the professor and the Charles D. Morgan/Acxiom Endowed Graduate Research Chair in Database and leads Social Awareness and Intelligent Learning (SAIL) Lab in Computer Science and Computer Engineering Department at the University of Arkansas. He got his BS degree in Information Science from the University of Science and Technology of China in 1994, ME degree in Computer Engineering from the Chinese Academy of Space Technology in 1997, and Ph.D. in Information Technology from George Mason University in 2001. Dr. Wu's major research interests include data mining, privacy and security, fairness aware learning, and big data analysis. Dr. Wu has published over 120 scholarly papers and served on editorial boards of six international journals and many program committees of top international conferences in data mining and AI. Dr. Wu is also a recipient of NSF CAREER Award (2006) and several paper awards including PAKDD'13 Best Application Paper Award, BIBM'13 Best Paper Award, CNS'19 Best Paper Award, and PAKDD'19 Most Influential Paper Award.

Dr. Valerie Satkoske

Valerie Satkoske is the Director of Ethics for Wheeling Hospital in Wheeling, West Virginia and serves as the Executive Director of the West Virginia Network of Ethics Committees and Associate Director of the West Virginia University Center for Health Law and Ethics. Dr. Satkoske is an assistant professor in the West Virginia University School of Medicine and is core faculty in the University of Pittsburgh Center for Bioethics and Health Law. She sits on the ethics committees of Wheeling Hospital, WVU Medicine—J.W. Ruby Memorial Hospital, and UPMC Passavant Hospital. Dr. Satkoske’s current research focuses on enhancing the effectiveness of communication between medical providers and decisionally-capacitated adults on the autism spectrum. She received her PhD in Healthcare Ethics from Duquesne University in Pittsburgh, Pennsylvania and her MSW from the University of Pittsburgh.

Dr. Yongkai Wu

Yongkai Wu is an upcoming Assistant Professor in the Department of Electrical and Computer Engineering at Clemson University. He earned his Ph.D. degree in Computer Science at the University of Arkansas in 2020 and B.Eng. degree in Electronic Engineering from Tsinghua University, China in 2014. His research interests focus on machine learning, data mining, and artificial intelligence, particularly fairness-aware machine learning and causal inference. His publications have appeared in prestigious conferences including IJCAI, KDD, NeurIPS, WWW, and a premier journal TKDE. He has served as a PC member for several international conferences, e.g. AAAI, IJCAI, KDD, NeurIPS, PAKDD.

Dr. Lu Zhang

Lu Zhang is an Assistant Professor in Computer Science and Computer Engineering Department at University of Arkansas. He received the BEng degree in computer science and engineering from University of Science and Technology of China, in 2008, and the PhD degree in computer science from Nanyang Technological University in 2013. His research interests lie in the field of data mining, machine learning, artificial intelligence, and distributed computing, particularly in discrimination/fairness-aware machine learning, causal modeling and inference, and data mining and privacy for genetic data.

[Day 2, Topic 6] AI in Smart Health: War On Opioids

This tutorial will consist of two parts. In the first part (titled "Probing into Illicit Drug Networks: a Data Mining Perspective"), we will first cover the theoretical foundation of heterogeneous information networks (HIN) and network embedding techniques such as DeepWalk, Node2Vec and Metapath2Vec. Then we will use darknet and surface net markets as examples to illustrate the problem and solution to illicit drug trade detection. In the second part (titled "Deep learning for Drug-Drug Interaction (DDI) prediction"), my PhD student and Graduate Research Assistant (Mr. Nathan Utzman) will present some background material related to DeepDDI (the first DNN-based tool for DDI prediction) and his recent work on leveraging this tool to opioid-related overdose death study.

Speaker(s)

Dr. Xin Li

Dr. Xin Li received the B.S. degree with highest honors in electronic engineering and information science from University of Science and Technology of China, Hefei, in 1996, and the Ph.D. degree in electrical engineering from Princeton University, Princeton, NJ, in 2000. He was a Member of Technical Staff with Sharp Laboratories of America, Camas, WA from Aug. 2000 to Dec. 2002. Since Jan. 2003, he has been a faculty member in Lane Department of Computer Science and Electrical Engineering. His research interests include image/video processing and computer vision. Dr. Li was elected a Fellow of IEEE in 2017.

Nathan Utzman

Nathan Utzman is a PhD student and Graduate Research Assistant at the Lane Department of Computer Science and Electrical Engineering, West Virginia University. His work focuses on applying artificial intelligence to drug-drug interactions and the opioid crisis in the United States. More generally, he is interested in medical applications of computer science.

[Day 2, Topic 7] Hands-on with Practical AI platforms

This session will focus on the use three platforms or frameworks for big data science and artificial intelligence. This will be a hands-on session and participants will examine and use the discussed frameworks to analyze large datasets and experiment with machine learning methodologies. As this is a hands-on session, participants will need to have access to a laptop or desktop computer (rather than a phone or Ipad). Also, since some of the activities will use tools from Google, each participant should have a Google account (such as gmail).

Speaker(s)

Don McLaughlin

Don McLaughlin is a project manager for the CRESH project and an instructor in the Lane Department of Computer Science and Electrical Engineering at West Virginia University. His teaching portfolio include data science, artificial intelligence and computer graphics. Formerly, he managed the high performance computer facility at WVU. He was also West Virginia University’s representative to Internet2 for a number of years.

For more information, please contact Don Adjeroh, PhD, Professor and Associate Chair of LCSEE, Project Lead PI.