Professor Hamido Fujita
Director, Intelligent Software Systems, Iwate Prefectural University, Japan
Biography: Hamido Fujita is professor at Iwate Prefectural University (IPU), Iwate, Japan, as a director of Intelligent Software Systems. He is the Editor-in-Chief of Knowledge-Based Systems, Elsevier of impact factor (4.528) for 2016. He received Doctor Honoris Causa from O’buda University in 2013 and also from Timisoara Technical University in 2018, and a title of Honorary Professor from O’buda University, Budapest, Hungary in 2011. He received honorary scholar award from University of Technology Sydney, Australia on 2012. He is Adjunct professor to Stockholm University, Sweden, University of Technology Sydney, National Taiwan Ocean University and others. He has supervised PhD students jointly with University of Laval, Quebec, Canada; University of Technology, Sydney, Australia; Oregon State University (Corvallis), University of Paris 1 Pantheon-Sorbonne, France and University of Genoa, Italy. He has four international Patents in Software System and Several research projects with Japanese industry and partners. He is vice president of International Society of Applied Intelligence, and Co-Editor in Chief of Applied Intelligence Journal, (Springer). He has given many keynotes in many prestigious international conferences on intelligent system and subjective intelligence. He headed a number of projects including Intelligent HCI, a project related to Mental Cloning as an intelligent user interface between human user and computers and SCOPE project on Virtual Doctor Systems for medical applications.
Title of Speech: Challenges on Big data based Clouds Health-Care for Risk Predictions based on Ensemble Classifiers and Subjective Analysis
Abstract: Discovering patterns from big data attracts a lot of attention due to its importance in discovering accurate patterns and features that are used in predictions of decision making. The challenges in big data analytics are the high dimensionality and complexity in data representation analytics especially for on-line feature selection. Granular computing and feature selection on data streams are among the challenge to deal with big data analytics that is used for Decision making. We will discuss these challenges in this talk and provide new projection on ensemble deep learning techniques for on-line health care risk prediction. Different type of data (time series, linguistic values, interval data, etc.) imposes some difficulties to data analytics due to preprocessing and normalization processes which are expensive and difficult when data sets are raw, or imbalanced. We will highlight these issues through project applied to health-care for elderly, by merging heterogeneous metrics from multi-sensing environment providing health care predictions assisting active aging at home. We have utilized ensemble learning as multi-classification techniques on multi-data streams using incremental learning to update data change “concept drift“ Subjectivity (i.e., service personalization) would be examined based on correlations between different contextual structures that are reflecting the framework of personal context, for example in nearest neighbor based correlation analysis fashion. Some of the attributes incompleteness also may lead to affect the approximation accuracy. I present deep learning feature selection in medical application early predictions (heart diseases and others). We outline issues on Virtual Doctor Systems, and highlights its innovation in interactions with elderly patients, also discuss these challenges in multiclass classification and decision support systems research domains. In this talk I will present the current state of art and focus it on health care risk analysis applications with examples from our experiments.
Professor Oleg Burdakov
Linkoping University, Sweden
Biography: Oleg Burdakov is a professor at Linkoping University, Sweden, since 1999, and also he is Affiliated Faculty at Center for Applied Optimization, University of Florida, Gainesville, USA. In 1998 he was a visiting professor at University of Campinas, Brazil. He was a research scientist at the Computing Center of the Russian Academy of Sciences (1980-1994) and CERFACS, Toulouse, France (1995-1997). He received Master (1977) and PhD Degre (1980) from the Moscow Institute of Physics and Technology. He is Editor-in-Chief of the journal Optimization Methods & Software. In 2014 he was among the winners of the International Implementation Challenge in solving Steiner tree problems organized by DIMACS, the US Center for Discrete Mathematics and Theoretical Computer Science which is a collaboration between Rutgers University, Princeton University, and the research firms ATT, Bell Labs, Applied Communication Sciences, and NEC. In 2017, he received a Visiting Scientist award under the Chinese Academy of Sciences President's International Fellowship Initiative. His research interests include: numerical methods for solving optimization problems and systems of nonlinear equations, in particular, Newton-type, stable secant and interpolation methods, globalization strategies; cardinality-constrained optimization; inverse problems; multilinear least-squares, nonsmooth optimization and equations; linear and nonlinear saddle problems and monotone equations; monotonic regression: data fitting and interpolation; hop-restricted shortest path and Steiner tree problems.
Title of Speech: A Bi-criteria Approach to Solving Huge Hop-constrained Steiner Tree Problems with Application to Positioning Unmanned Aerial Vehicles as Communication Relays
Abstract: We consider the directed Steiner tree problem (DSTP) with a constraint on the total number of arcs (hops) in the tree. This problem is known to be NP-hard. Only heuristics can be applied in the case of its instances whose size is beyond the capacity of the existing exact algorithms. The hop-constrained DSTP is viewed as a bi-criteria problem in which the tree cost and the number of hops are minimized. We derive optimality conditions and use them for developing an approach aimed at approximately solving hop-constrained DSTP. The approach can also be used for improving approximate solutions produced by other heuristic algorithms or as a part of exact algorithms. Specific label-correcting-type algorithms based on this approach will be presented, and preliminary results of their performance on a set of test problems will be reported. The test instances originate from 3D placement of unmanned aerial vehicles used for multi-target surveillance. They are characterized by a relatively small number of terminal nodes and a very large number of nodes and a huge number of arcs (over one billin).
Professor Andries P. Engelbrecht
Stellenbosch University, South Africa
Biography: Prof Andries Engelbrecht received the Masters and PhD degrees in Computer Science from the University of Stellenbosch, South Africa, in 1994 and 1999 respectively. He is Voigt Chair in Data Science in the Department of Industrial Engineering, with a joint appointment as Professor in the Computer Science Division, Stellenbosch University. His research interests include swarm intelligence, evolutionary computation, artificial neural networks, artificial immune systems, and the application of these Computational Intelligence paradigms to data analytics, games, bioinformatics, finance, and difficult optimization problems. He is author of two books, Computational Intelligence: An Introduction and Fundamentals of Computational Swarm Intelligence
Title of Speech: A Hyper-heuristic Framework for Real-Valued Dynamic Optimization
Abstract: Dynamic optimization problems provide a number of challenges to optimization algorithms, requiring algorithms to find and track one or more optima in a search landscape that changes over time. The difficulty in tracking optima increases with increase in severity and frequencies of changes in the search landscape. In addition, it may happen that existing optima disappear, or that new ones appear. In order to be successful in this task, optimization algorithms have to employ efficient strategies to maintain exploration, while still keeping sufficient exploitation capabilities. Various meta-heuristics have been developed to solve dynamic optimization problems, to various degrees of success. Algorithms that are efficient in progressive environments may perform poor in abrupt and chaotic environments, and vice versa. Without prior knowledge about severity and frequency of changes, and if changes are random or follow some pattern, it is very difficult to decide on which algorithm to use. Hyper-heuristics offer a solution to this dilemma, in that a dual optimization process is executed. During the search of an optimum, a hyper-heuristic makes use of a heuristic selection operator to "learn" which meta-heuristic form a pool of meta-heuristics is best, and any point in time, to solve the problem. While much work has been done in the development of hyper-heuristics, most of the research efforts focussed on static optimization problems. This talk will present a hyper-heuristic framework for real-valued dynamic optimization, where a pool of dynamic meta-heuristics are evaluated by a selection operator in order to find the best meta-heuristic to find and track an optimum in a dynamic environment, without any prior knowledge about the severity and frequency of changes in the search landscape.