Way to Cheongju
Way to Cheongju from Incheon Airport
By limousine Bus Take a limousine bus for Cheongju at platform no. 8 of the first floor in terminal 1, and platform no 9 of the basement 1 floor in the terminal 2. Bus fair is 21,600won. You may purchase the bus ticket at ticket offices located on the first floor in terminal 1 and on the basement 1 floor in terminal 2. Duration is about 2 hours. Get off at Cheongju Bus Terminal and take a taxi to Chungbuk National University. (https://www.airport.kr/ap/en/tpt/busRouteList.do) [Terminal 1]





Way to Chungbuk National University, Korea
For more information, please visit:
Design and optimization of Supply Chain Network (SCN), Reverse Logistics and Sustainable Logistics is very important issue, which plans, implements and controls the efficient and effective forward/reverse flows and storage of goods, services and related information between the point of origin and the consumption to meet customers’ requirements. In the real world of such logistics systems, many combinatorial optimization problems (COPs) impose on more complex issues, such as complex structure, nonlinear constraints, and multiple objectives to be handled simultaneously and make the problem intractable to the traditional approaches because of NP-hard COP. In order to develop an efficient algorithm that is in a sense “best solution”, i.e., whose reasonable computational time for NP-hard COPs met in practice, we have to consider the following very important issues: Quality of solution, Computational time and Effectiveness for multi-objective COPs. Metaheuristics including Evolutionary Algorithm (EA) is a generic population-based Bio-inspired Computation such as Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Cuckoo Search (CS) and Teaching-Leaning Based Optimization (TLBO) algorithms. EA is based on principles from evolution theory, and it is very powerful and broadly applicable stochastic search and heuristic optimization which is effective for solving various NP hard COP models.
This tutorial talk will be summarized recent advances of several metaheuristics such as Hybrid GA (HGA) with Fuzzy Logic Controller and Local search, Hybrid GA with PSO and Hybrid GA, PSO with TLBO for applying to various scheduling and routing problems. Secondly, based on hybrid metaheuristics, real applications such as intelligent manufacturing, semiconductor manufacturing, integrated SCN, and sustainable logistics models will be summarized.
As deep neural network techniques become more advanced, there is growing demand for edge AI in mobile or IOT applications. Conventional computation hardware for neural networks heavily rely on massive parallel programmable architectures such as GPUs. While such architectures offer high flexibility and programmability, they often fail to satisfy the requirements of compact size and low energy operation for edge AI. Recent research has reveled that more efficient accelerator architectures can save power consumption in an order of magnitude for CNN inference with smaller silicon size. This talk covers a few recent achievements in accelerator design and optimization research including a CNN accelerator based on diagonal cyclic array for minimal memory access, mixed signal convolutional kernel accelerator to overcome the inherent complexity of digital counterpart, and optimal floating-point datapath for compact accelerator of CNN training.
Deep learning (DL) has been widely investigated and applied for many engineering applications. Broad learning system (BLS) has been shown to work as an effective and efficient incremental learning without the need for deep architecture, thus giving a new paradigm and learning system for AI systems. By incorporating with the merits of DL, variant BLSs and fuzzy logics, this talk will present you fuzzy DL‐based and fuzzy BLS‐based control frameworks for a class of nonlinear systems and mobile robots and multirobots. In the short talk, some advances on fuzzy DL NN and fuzzy BLSs are first reported, their applications to control of nonlinear dynamic systems and wheeled robots and multirobots are mentioned in some detail. Experimental results are provided to illustrate the merits of the proposed methods. Last but not least, some perspective topics on fuzzy deep and broad learning methods are recommended for future research.
What is intelligence? What is the nature/essence/quiddity of intelligence? Can intelligence be defined as a comprehensive/universal and unique concept? In psychology and related fields, intelligence is a fundamental concept in the theory and analysis of human behavior. On the other hand, intelligence is a fundamental concept in the design and construction of any intelligent system. In fact, by any definition of intelligence, we achieve a different intelligent system.
In this talk, based on the definitions proposed by two categories of scientists (1: psychologists, and 2: intelligent systems specialists), we identify the main components of such definitions. After that, based on these components, we categorize the types of definitions of intelligence. Then, we try to present a comprehensive definition of intelligence, so that it encompasses all aspects of intelligence. We will see the proposed definition has a strong connection with the notion of uncertainty. As such, the connection between intelligence and uncertainty becomes more clear, and as a result, it paves the way forward for human behavior analysts and intelligent systems specialists.
“Human Symbiotic Systems” (HSS) aims at studying the basic principles and methods of designing intelligent interaction in bidirectional communication based on the effective collaboration and symbiosis between humans and artifacts such as robots, agents, and computers. The research on HSS includes a wide range of topics from a variety of fields such as human-agent interaction, human-machine interface, intelligent robotics, Kansei engineering and so on. I established a special interest group on Human Symbiotic Systems in Japan Society for Fuzzy Theory and Intelligent Informatics (SOFT) in 2007 in order to
advance researches in this field.
The soft computing method is a technology that bridges the gap between “intelligent systems” which are based on logic and regularity, and “humans” who have ambiguity and flexibility, and is also a core technology for research on human symbiotic systems. Both “human intelligence” and “artificial intelligence” are superior, but it is necessary to build efficient cooperative systems between them in order to achieve excellent human symbiotic systems in the future. In this talk, I will give an overview of the activities of HSS research group in SOFT, and introduce some of my own previous researches related to human symbiotic systems.
Fuzzy relational compositions are by far not a recent concept, as a generalization of compositions of classical binary relations they have been studied in late 1970’s and early 1980’s mainly by Wyllis Bandler and Ladislav J. Kohout and let on followed by many other scholars, e.g., Etienne E. Kerre, Witold Pedrycz, Radim Belohlavek, or Bernard De Baets. The huge potential that was given to the society was, unfortunately, partly unused, due to some application limitations of the compositions. Indeed, for example, the standard compositions are based either on the existential quantifier or on the universal quantifier and there is a huge gap between these two. Another problem is that a single composition cannot capture all what needs to be captured for particular application needs. However, if we dare to enrich the existing knowledge of fuzzy relational compositions by other tools such as generalized quantifiers filling the gap between the existential one and the universal one, and we consider the compositions to be only basic bricks, we have a key to unlock the door to an exciting journey through the expressive power of fuzzy relational constructions built from combinations of simple yet powerful bricks.
In the recent past, big data analytics has added some new challenges to the conventionally used feature selection methods. This is because there are some unique characteristics of big data that have made researchers think differently when dealing with the feature selection problem. For example, big data not only include a vast amount of redundant and non-informative features, but also have noises of varied degrees and types, which significantly augments the difficulty of selecting relevant features. Some data are also unreliable, due to different means of data acquisition, which further enhances the complexity of feature selection. In this talk, we will explore the challenges related to feature selection for big data analytics from various aspects. From the perspective of data, we will discuss structured features, linked data, multi-source data and streaming data. Besides, from the performance perspective, we will discuss two challenges – scalability and stability.
Abstract: Spiking neural networks, usually considered the third generation of artificial neural networks, consist of more biologically plausible artificial neurons which receive and send out spiking signals. They have attracted much attention from both neuroscience and engineering due to their biological plausibility and energy efficiency. Their behaviors are affected by their neurons’ inherent characteristics, network topology, and spike-based data encoding methods. This diversity makes it difficult to train spiking neural networks. Various training algorithms have been developed for spiking neural networks. They can be categorized into local training, direct training, ANN-SNN conversion, and hybrid training. This lecture addresses the characteristics and underlying assumptions of some important training algorithms for each of these categories. It also discusses the pros and cons of spiking neural networks in terms of software and hardware implementations.