KEYNOTE

Keynote Speech 1: Why Rectified Linear Neurons: Two Convexity-Related Explanations

Vladik Kreinovich, Vice President of IFSA, University of Texas at El Paso, USA

Biography: Vladik Kreinovich received his MS in Mathematics and Computer Science from St. Petersburg University, Russia, in 1974, and Ph.D. from the Institute of Mathematics, Soviet Academy of Sciences, Novosibirsk, in 1979. From 1975 to 1980, he worked with the Soviet Academy of Sciences; during this time, he worked with the Special Astrophysical Observatory (focusing on the representation and processing of uncertainty in radioastronomy). For most of the 1980s, he worked on error estimation and intelligent information processing for the National Institute for Electrical Measuring Instruments, Russia. In 1989, he was a visiting scholar at Stanford University. Since 1990, he has worked in the Department of Computer Science at the University of Texas at El Paso. In addition, he has served as an invited professor in Paris (University of Paris VI), France; Hannover, Germany; Hong Kong; St. Petersburg and Kazan, Russia; and Brazil.

His main interests are the representation and processing of uncertainty, especially interval computations and intelligent control. He has published eight books, 24 edited books, and more than 1,500 papers. Vladik is a member of the editorial board of the international journal “Reliable Computing” (formerly “Interval Computations”) and several other journals. In addition, he is the co-maintainer of the international Web site on interval computations http://www.cs.utep.edu/interval-comp.

Vladik is Vice President of the International Fuzzy Systems Association (IFSA), Vice President of the European Society for Fuzzy Logic and Technology (EUSFLAT), Fellow of International Fuzzy Systems Association (IFSA), Fellow of Mexican Society for Artificial Intelligence (SMIA), Fellow of the Russian Association for Fuzzy Systems and Soft Computing; he served as Vice President for Publications of IEEE Systems, Man, and Cybernetics Society 2015-18, and as President of the North American Fuzzy Information Processing Society 2012-14; is a foreign member of the Russian Academy of Metrological Sciences; was the recipient of the 2003 El Paso Energy Foundation Faculty Achievement Award for Research awarded by the University of Texas at El Paso; and was a co-recipient of the 2005 Star Award from the University of Texas System.

Keynote Speech 2:

Modelling and Constructing Intelligent Forecasting System for Agricultural Applications

Liya Ding, School of Science and Technology, Meiji University, Japan

Abstract

Artificial intelligence (AI) and machine learning (ML) have been applied to intelligent systems in broad areas – from industry, agriculture, and medicine to education, and business – in almost all fields where humans carry out intelligent decision-making activities. Prediction and classification are among the major tasks required in AI and ML applications. The latest developments of AI and ML offer more powerful algorithms and models, and more useful tools that provide convenience to application developers. While intelligent algorithms and learning models play the key role, reasonable problem modelling and the availability of data are essential for the success of an intelligent system. This talk takes two application cases – frost forecasting and greenhouse gas (CH4) forecasting – to discuss the challenges in system modelling and construction and explore possible strategies for problem solving.

The first application case discussed is on frost forecasting. Frost is a climate phenomenon that can happen anywhere in the world, causing damage to crop plants. To effectively protect plants from frost damage, an early alarm of frost can be helpful for growers. Frost is a localized phenomenon and can be quite variable across a small area, so predictive models developed using local data with good resolution are preferred. The occurrence of frost is closely related to multiple environmental factors, including temperature, humidity, radiation and more. Furthermore, the occurrence of frost at a specific moment is caused by the prior movement of environmental factors. This means there is some cause-and-effect between environmental factors and the occurrence of frost, governed by a process taking place in time. Current ML captures or discovers an association relation of factors from data, but not a cause-and-effect. Therefore, modelling strategies as compensation are needed before adopting existing ML models. Main challenges are from the unclear time delay of the impact of environmental factors, incomplete set of features with limited sensor data, and data labelling of frost as it is not directly measurable with sensor equipment. Modelling strategies that will be presented include causal modelling with time delay, variable granulation, ensemble modelling with different time settings, and a hybrid system with time series forecasting of environmental factors and predictive models. Auto-labelling strategy will also be discussed.

The second application case discussed is on greenhouse gas forecasting. Methane (CH4) is a hydrocarbon that is a primary component of natural gas. Methane is also a greenhouse gas (GHG), so its presence in the atmosphere affects the earth’s temperature and climate system. Practitioners put in efforts to preventing GHG from happening, typically through controlling the water level of rice paddies in the case of CH4. Air temperature and soil water content are considered highly relevant to CH4, among other factors. Data collection for CH4 forecasting is more difficult in comparison to frost forecasting as soil water content can be changed directly by human action. Another challenge is that CH4 occurrence takes more time, so daily changes in air temperature becomes less relevant. Strategies for the decomposition of time series data to capture past trends, and algorithms to predict future CH4 levels if no action is taken will be presented. Such estimates can serve as a useful reference for possible water management action.

Through these two application cases, the importance of problem modelling for the construction of intelligent systems based on domain understanding and knowledge will be emphasized. For system implementation, existing predictive models and time series forecasting models are adopted. Experiment results will be provided.

Biography

Liya Ding received her B.E. degree in Computer Engineering from Shanghai University of Technology (now Shanghai University), China in January 1982, and her Ph.D. degree in Computer Science with the research on Fuzzy Prolog from Meiji University, Japan in March 1991.

She was at the Institute of Systems Science (ISS), National University of Singapore (NUS), Singapore from 1991 to 2002, where she served as a project leader for research programs on artificial intelligence, and neural networks and fuzzy logic (1991-1998); senior lecturer, senior consultant (1998-2002) and programme director (2000-2002) for Knowledge Engineering programme. She was also a key researcher (1993-1998) and the project leader of the Neuro-ISS Laboratory (1994-1996) for the Real World Computing Partnership (RWCP, an international project supported by the Ministry of International Trade and Industry, Japan 1992-2002).

From 2002 to 2005, she was an associate professor and the dean of the School of Intelligent Systems and Technology in the Inter-University Institute of Macau (now the University of Saint Joseph), Macau. From 2005 to 2011, she was a full professor and the vice dean (2005-2008) then dean (2008-2010) of the Faculty of Information Technology, Macau University of Science and Technology, Macau. She rejoined ISS, NUS, in 2011 to 2017 as a member of the Knowledge Engineering programme.

From 2017 onwards, she has been serving as a full professor in the Department of Electronics and Bioinformatics, School of Science and Technology, Meiji University where she leads the Knowledge Engineering laboratory.

Her research interests are in Machine Learning Applications, Knowledge-Based Systems (KBS), Fuzzy Logic, Approximate Reasoning, and Soft Computing. In recent years, she has been working on intelligent systems in agricultural and industrial applications, using machine learning and intelligent techniques and exploring causal modelling to capture cause-and-effect relations with time delays and imperfect knowledge. She has published more than a hundred research papers in peer-reviewed journals and international conferences. She was the editor and contributor of the book “A New Paradigm of Knowledge Engineering by Soft Computing,” World Scientific, 2001. She also contributed a chapter to the memorial book for fuzzy logic, “Fifty Years of Fuzzy Logic and its Applications,” Springer, 2015, in which the first chapter is from late Professor Lotfi. A. Zadeh.

She proposed Knowware System (KWS) in 2006 for the modelling and constructing of intelligent systems and has received grants from the Macao Science and Technology Development Fund twice during 2007-2013 for the research on KWS as the principal investigator. She also received research grants from the Meiji University Graduate School in 2019 as the principal investigator for the “Development of Frost Forecast System by Machine Learning”, and from a Japanese water purification company in 2019 to 2021 for a joint research project in developing an intelligent decision support system for the optimal scheduling of equipment maintenance with a hybrid system using machine learning, fuzzy inference and KBS.

Liya Ding has served as an associate editor, member of editorial board, and reviewer for international journals since 1994; as a session chair, member of program committee and international advisory committee for numerous international conferences since 1992; and as the conference programme chair for the 19th International Conference on Knowledge Based and Intelligent Information and Engineering Systems (KES2015) in Singapore in 2015.