AI and IoT:
Acquiring Knowledge in Resource-Constrained Environments

The Internet of Things and Artificial Intelligence go hand in hand: How should you make sense of the petabytes of data arriving from IoT applications, if not with AI? And what could be more interesting data for AI than those arriving from IoT nodes, measuring what is going on in the real world? Yet there are more connections between these two areas, which we will explore in this seminar.

We present five speakers to discuss IoT and AI, and discover more connections between these two areas. The seminar is open to everyone, and will be interesting to a wide audience from many disciplines. The time is Thursday June 22, 2017. 9:00 to 15:00, KJEL2, Gløshaugen.

Slides are available now. Just follow the links in the programme below.

Programme

09:00 Welcome and Introduction
09:15 Yih-Fang Huang: Towards Smarter IoT with Intelligent Sensors – Event-Triggered Sensing, Information Processing and Learning
10:15 Anthony Kuh: Signal Processing and Machine Learning Applied to Model Selection and Bad Data Detection for the Electrical Grid
11:00 Roberto Minerva: Optimization of Communications Resources in Large IoT Systems
12:00 Lunch
13:00 Jo Eidsvik: Value of Information Analysis in Models with Statistical Dependence
14:00 Kerstin Bach: Collaborative Research at the NTNU-Telenor AI Lab

Speakers and Talks

Towards Smarter IoT with Intelligent Sensors – Event-Triggered Sensing, Information Processing and Learning

Yih-Fang Huang, Professor of Electrical Engineering, University of Notre Dame, IEEE Fellow

Biography 
 

The continued evolution of technologies in sensing, communications, networking and information processing has made possible the advent of “Internet of Things” (IoT). In essence, IoT is part of the continued evolution and expansion of the digital world and it represents one of the new frontiers. It has been noted that IoT will play an important role in the development of smart cities, management and monitoring of (large-scale) smart infrastructures that include smart grid, transportation systems, water pipelines, etc., building and home automation, hospital and healthcare facilities, and environmental monitoring. Read more...

Signal Processing and Machine Learning Applied to Model Selection and Bad Data Detection for the Electrical Grid

Anthony Kuh, University of Hawaiʻi. IEEE Fellow. Program Director at the National Science Foundation

Biography 
 

This talk focuses on using signal processing and machine learning applied to renewable energy and the smart grid. We start off by giving an introduction to the Hawai`i energy landscape and the University of Hawai`i. Read more...

Optimization of Communications Resources in Large IoT Systems

T.B.A.

Value of Information Analysis in Models with Statistical Dependence

We constantly use information to make decisions about utilizing and managing resources. Decision situations are often complex, involving multivariable interactions of uncertainties. How can we quantitatively analyze and evaluate different information sources in this context? What is the value of data and how much data is enough?

The value of information is a concept in decision theory for analyzing the value of obtaining additional information to solve a problem. The value of information is computed before gathering data, and can be useful for checking if data acquisition or processing is worthwhile doing, or for comparing various test opportunities. We will distinguish between perfect versus imperfect information, and total versus partial information where only a subset of the data is acquired or processed. Examples demonstrate value of information analysis in various applications, including borehole data in mining, seismic data in petroleum, and autonomous surveying by submarines for environmental mapping.

Collaborative Research at the NTNU-Telenor AI Lab

Kerstin Bach is an Associate Professor at the Department of Computer Science at NTNU. Her research interests are intelligent decision support systems, data analysis and knowledge-based systems, particularly case-based reasoning. She teaches machine learning and advanced knowledge-based methods at NTNU and supervises students in the field since 2007. Her genuine interest in machine learning techniques and their application lead into participation in various EU and industry funded projects at the German Research Center for Artificial Intelligence as well as NTNU. Currently, among other research activities, she is the project manager of the selfBACK EU project that develops an eHealth application for low back pain patients. She enjoys discussing AI methodologies and applications with fellow researchers, but more importantly teaching it to students and bringing it into multidisciplinary projects and thereby shaping the new digital world.

Towards Smarter IoT with Intelligent Sensors - Event-Triggered Sensing, Information Processing and Learning

Yih-Fang Huang

The continued evolution of technologies in sensing, communications, networking and information processing has made possible the advent of “Internet of Things” (IoT). In essence, IoT is part of the continued evolution and expansion of the digital world and it represents one of the new frontiers.  It has been noted that IoT will play an important role in the development of smart cities, management and monitoring of (large-scale) smart infrastructures that include smart grid, transportation systems, water pipelines, etc., building and home automation, hospital and healthcare facilities, and environmental monitoring. 

Sensor technologies are at the forefront of IoT, for sensors are the “things” that take physical measurements and convert them into data from which information and intelligence are obtained.  Intelligent sensors, coupled with effective machine learning algorithms, can significantly enhance the system’s ability to quickly and accurately extract information from collected data and convert the information into timely actionable intelligence.  Intelligent sensors can be made more autonomous – they can be more discerning in their actions of taking measurements, processing data and transmitting information.  Consequently, intelligent sensors are capable of separating irrelevant data from relevant data, thereby facilitating timely and accurate actions.  More importantly, intelligent sensors are energy-savvy due to their more discerning actions.

This presentation will begin with discussions on benefits of intelligent sensors as they relate to IoT.  We shall show that event-triggered (or event-driven) sensing, processing and transmission is a viable approach to designing intelligent sensors.  We will then use an example of distributed adaptive signal processing that implements event-triggered sensing, information processing and learning.   In particular, we shall show an event-triggered diffusion distributed adaptive estimation paradigm that can save on the overhead of communication and signal processing without significantly compromising the system performance.  Sensors designed according to such principles can clearly use the saved energy to perform other functions and/or render longer-lasting and more secure network.  Some simulation results will be given to illustrate the benefits of the event-triggered approach.

About Yih-Fang Huang

Dr. Yih-Fang Huang is Professor of Electrical Engineering and Senior Associate Dean for Education and Undergraduate Programs in the College of Engineering at the University of Notre Dame, Notre Dame, Indiana, U.S.A.  He received his B.S.E.E. degree from National Taiwan University in Taipei, Taiwan, M.S.E.E. degree from University of Notre Dame, M.A. and Ph.D. from Princeton University in Princeton, New Jersey.  From 1998 to 2006, he served as chair of Notre Dame’s Electrical Engineering department.

Dr. Huang’s research work employs principles in mathematical statistics to solve detection and estimation problems that arise in various applications, including wireless communications, distributed sensor networks, smart electric power grid, etc.  He has published more than 200 papers in archival journals and conference proceedings in those areas.

Dr. Huang is a Fellow of the Institute of Electrical and Electronic Engineers (IEEE) (’95).  He received the Golden Jubilee Medal of the IEEE Circuits and Systems Society in 1999, served as Vice President in 1997-98 and was a Distinguished Lecturer for the same society in 2000-2001.  He was the lead Guest Editor for a Special Issue on Signal Processing in Smart Electric Power Grid of the IEEE Journal of Selected Topics in Signal Processing, December 2014.  At the University of Notre Dame, he received Presidential Award in 2003, the Electrical Engineering department’s Outstanding Teacher Award in 1994 and in 2011, the Rev. Edmund P. Joyce, CSC Award for Excellence in Undergraduate Teaching in 2011, and the Engineering College’s Outstanding Teacher of the Year Award in 2013.  He has served on the Illinois Board of Higher Education in a role of chair of a review team for the Department of Electrical and Computer Engineering of University of Illinois, Chicago.  Recently, he also served as a member of an external review team for Illinois Institute of Technology, Department of Electrical and Computer Engineering.

In Spring 1993, Dr. Huang received the Toshiba Fellowship and was Toshiba Visiting Professor at Waseda University, Tokyo, Japan.  From April to July 2007, he was a visiting professor at the Munich University of Technology, Germany.  In Fall, 2007, Dr. Huang was awarded the Fulbright-Nokia scholarship for lectures/research at Helsinki University of Technology in Finland.  Dr. Huang was appointed Honorary Professor in the College of Electrical Engineering and Computer Science at National Chiao-Tung University, Hsinchu, Taiwan, in 2014.
 

Value of Information Analysis in Models with Statistical Dependence

Jo Eidsvik

We constantly use information to make decisions about utilizing and managing resources. Decision situations are often complex, involving multivariable interactions of uncertainties. How can we quantitatively analyze and evaluate different information sources in this context? What is the value of data and how much data is enough?

The value of information is a concept in decision theory for analyzing the value of obtaining additional information to solve a problem. The value of information is computed before gathering data, and can be useful for checking if data acquisition or processing is worthwhile doing, or for comparing various test opportunities. We will distinguish between perfect versus imperfect information, and total versus partial information where only a subset of the data is acquired or processed. Examples demonstrate value of information analysis in various applications, including borehole data in mining, seismic data in petroleum, and autonomous surveying by submarines for environmental mapping.

About Jo Eidsvik

Eidsvik has an MSc in applied mathematics from the University of Oslo and a PhD in statistics from NTNU. He has industry work experience from the Norwegian Defense Research Estalishment and from Statoil. His research interests are in spatio-temporal statistics and computational statistics, and in design of experiments and value of information analysis. He co-authored the book Value of Information in the Earth Sciences.

Signal Processing and Machine Learning Applied to Model Selection and Bad Data Detection for the Electrical Grid

Anthony Kuh

This talk focuses on using signal processing and machine learning applied to renewable energy and the smart grid. We start off by giving an introduction to the Hawai`i energy landscape and the University of Hawai`i.

We discuss modeling distributed solar PV energy sources. With higher penetrations of distributed solar PV energy sources new methods are needed to effectively model these distributed energy sources. These generally involve using more distributed state estimation methods modeling energy sources and loads using graphical approaches. Distributed state estimation approaches include using message passing algorithms such as the Belief Propagation Algorithm (BPA). Here we look at approximations of the distributed energy sources using tree structures. We look at existing algorithms such as the Chow-Liu tree approximation algorithm using the Kullback Leibler (KL) Divergence and discuss the quality of approximation algorithms by formulating the problem as a detection problem and considering Receiver Operating Curves (ROC)s and the Area Under the Curve (AUC). A key quantity that we define is the Correlation Approximation Matrix (CAM). The KL Divergence, Jeffreys Divergence, and AUC all depend directly on the eigenvalues of the CAM. Simulations on real and simulated data show the quality of the tree approximations.

We then discuss detecting bad data for the electrical grid. We use a machine learning approach by formulating an online sparse one-class least squares support vector machine (OC)-(LS)-(SVM). The online OC-LS-SVM achieves sparsity by using the approximate linear dependence criterion and detects outliers by classifying data using a threshold test. We then test our algorithm on IEEE bus simulation data. We inject bad data at critical locations, inject multiple bad data, and use false data injection attacks. The online OC-LS-SVM performs better on all tests than traditional state estimation methods using the largest residual test method.

About Anthony Kuh

Anthony Kuh received his B.S. in Electrical Engineering and Computer Science at the University of  California, Berkeley in 1979, an M.S. in Electrical Engineering from Stanford University in 1980, and  a Ph.D. in Electrical Engineering from Princeton University in 1987.  Dr. Kuh previously worked at AT&T Bell Laboratories and has been on the faculty in Electrical Engineering at the University of Hawai’i since 1986.   He is currently a Professor in the Department, serving as director of the interdisciplinary renewable energy and island sustainability (REIS) group, and is also serving as a program director for the National Science Foundation (NSF). Previously, he served as Department Chair of Electrical Engineering Dr.  Kuh's research is in the area of neural networks and machine learning, adaptive signal processing, sensor networks, communication networks, and renewable energy and smart grid applications.

Dr. Kuh won a National Science Foundation Presidential Young Investigator Award and is an IEEE Fellow.   He was also a recipient of  the Boeing A. D. Welliver Fellowship and received a Distinguished Fulbright Scholar’s Award working at Imperial College in London.  Dr. Kuh was an Associate Editor for the IEEE Transactions on Circuits and Systems, served on the IEEE Neural Networks Administrative Committee,  served on the IEEE Neural Networks for Signal Processing Committee, and was a Distinguished Lecturer for the IEEE Circuits and Systems Society. Dr. Kuh co-chaired the 1993 International Symposium on Nonlinear Theory and its Applications (NOLTA) and served as the technical co-chair for the 2007  IEEE ICASSP both held in Honolulu. He served as the IEEE Signal Processing Society Regions 1-6 Director at Large and was was a senior editor of the IEEE Journal of Selected Topics in Signal Processing. He currently serves on the Board of Governors of the Asia Pacific Signal and Information Processing Association as Vice President of Technical Activities.

In January, 2017 he started service as a program director for NSF. He is in the Electrical, Communications, and Cyber Systems (ECCS) division working in the Energy, Power, Control, and Network (EPCN) group.