Neumann János Egyetem GAMF Műszaki és Informatikai Kar

Gyengénlátó Változat

Final Program

PDF letöltés Cikk nyomtatása E-mail

8:30 – 9:00 Registration (Building 4, 3rd Floor, Room 311)

9:00 – 9:05 Welcome Address: Dr. Zoltán Weltsch, vice rector for research

Session Chair: Dr. Tamás Kovács

9:05 – 9:25 Radu-Emil Precup: Mechatronics Applications of Evolving Takagi-Sugeno-Kang Fuzzy Models

Abstract: Evolving Takagi-Sugeno-Kang fuzzy models are characterized by continuous online rule base learning. The structure and parameters of these fuzzy models are computed by online identification algorithms, which continuously evolve the parameters of fuzzy models. Fuzzy models are constructed online in terms of the adding mechanism that adds new local models or removes old ones.

This presentation is focused on certain applications concerning the development of evolving Takagi-Sugeno-Kang fuzzy models for four mechatronics applications: finger dynamics for prosthetic hand myoelectric-based control, twin rotor aerodynamic systems, anti-lock braking systems, and vehicles with continuously variable transmission systems. The results are obtained by the Process Control group of the Politehnica University of Timisoara, Romania. Simulation and experimental results are outlined.

9:25 - 9:45 László Pokorádi: Interval Uncertainty Analysis of Systems' with Complex Interconnections Reliability

Abstract:  Industrial reliability and safety are important phenomena. The uncertainties play a key role in quantitative investigation of reliability of Systems with Complex Interconnections (SwCI). This paper proposes a modular approach matrix algebraic interval analysis method to investigate uncertainty of Bridge Structure Systems (BSS) reliability. The uncertainty intervals of system reliability can be determined by the proposed method. Using these results maintenance cost and working expenditures of investigated manufacturing unit can be estimated, the maintenance management can receive specific supporting data to make correct decisions.

9:45 – 10:05 Annamária Koncz: Safety related automotive systems based on fuzzy logic

Abstract: Due to nowadays fast technological changes, new safety funcions need to be implemented in vehicles systems as well. Faster and more complex vehicle systems require more intelligent safety solutions.
This presentations summarizes safety related components of vehicle systems, and the future of these applications.
In case of safety related automotive systems the application of fuzzy logic is an indisputable element. In this case the focus is the link between fuzzy logic and driving safety.

10:05 – 10:25 János Botzheim: Evolutionary and Swarm Optimization Algorithms for Hyperparameter Optimization of Machine Learning Models 

Abstract: In the field of machine learning we usually apply neural networks with various structure. The algorithms of the deep learning techniques and the structure of the applied networks have several parameters that have a huge impact on the performance of the search technique. These parameters are called hyperparameters. The aim of our current research is to optimize these hyperparameters using evolutionary and swarm based optimization algorithms.

10:25 – 10:45 Coffee break

Session Chair: Dr. Attila Pásztor

10:45 - 11:05 Tamás Tompa: Q-learning vs. FRIQ-learning in the Maze problem

Abstract: The goal of this presentation is to give a demonstrative example for introducing the benefits of the FRIQ-learning (Fuzzy Rule Interpolation-based Q-learning) versus the traditional discrete Q-learning. The chosen example is an easily scalable discrete state and discrete action space task the Maze problem. The main difference of the two studied reinforcement learning methods, that the traditional Q-learning has discrete state, action and Q-function representation. While the FRIQ-learning has continuous state, action space and a Fuzzy Rule Interpolation based Q-function representation. For comparing the convergence speed of the two methods, both will start from an empty knowledge base, zero Q-table for the Q-learning and empty rule-base for the FRIQ-learning and following the same policy stops at the same performance condition. In the example of the paper the Maze problem will be studied in different obstacle configurations and different scaling.

 11:05-11:25 Sara Imene Boucetta: Software Defined Networks Application in VANETs

Abstract: The concept of ambient intelligence is to create an intelligent and immediate user space integrated to the environment. One of its applications is Intelligent Transportation System (ITS), which aims to improve the safety, efficiency and conviviality of road transports. Vehicles travel at high speeds and it is not possible to establish an infrastructural network between them and they are committed to follow a precise route during their movements and their high-speed movements generate a high rate of connection and disconnection in the network. This problem resulted in the birth of Vehicular Ad-hoc Networks (VANETs) that allow communication among vehicles (V2V) and between vehicles and fixed infrastructure (V2I), aiming to support a wide range of services and applications. However, there are some serious technical challenges in VANET communication which are mainly related to the high dynamicity and volatility of the vehicular environment. To overcome these challenges a promising option is the application of Software Defined Networking (SDN). It is a dynamic, manageable, cost-effective, and adaptable architecture. SDN decouples network control and forwarding functions, enabling network control to be directly programmable and the underlying infrastructure to be abstracted from applications and network services. The flexibility of SDN makes it an approach that can be used to satisfy the requirements of VANET scenarios and to bring the programmability and flexibility to this distributed wireless network, while simplifying its management and enabling new V2V and V2I services.

11:25 – 11:45 Péter András Agg, Zsolt Csaba Johanyák: Energy Efficiency in SDN Networks

Abstract: A significant advantage of Software-Defined Networks (SDNs) over traditional networks is the ability to directly control network behavior and direct communication with network elements. These options are useful and provide us with a quicker, more effective response in critical situations. Of course, they do not relieve us of the task of designing and operating the network optimally. In my presentation, I will present some possible solutions for optimizing SDN networks, primarily in terms of energy management.

11:45-12:05 Szilveszter Kovács: FBDL: A Declarative Language for Behavior Modeling

Abstract: Behavior-based system (BBS) is a structure built upon behavior components, behavior coordination and behavior fusion. The goal of this presentation is to introduce a declarative language especially designed for supporting embedded behavior based applications. The suggested language adapts the BBS implementation based on fuzzy rule-based systems, fuzzy rule interpolation (FRI) and fuzzy state machines. This case the description of the BBS as a network of behaviors turns to a network of fuzzy systems. The suggested Fuzzy Behavior Description Language (FBDL), is a declarative language for defining FRI rule-bases and a network of fuzzy systems in a unified framework. According to the suggested concept, the FBDL code is “running”, as a parameter configuration, directly on a FRI state machine called “FRI Behavior Engine”. In an embedded behavior-based application, the FRI Behavior Engine with the parameter configuration fetched from the FBDL code can directly serve the actions according to the observations and the state of the system.

 

12:05-13:00 Lunch

Session Chair: Radu-Emil Precup

13:00-13:20 László Kovács: Incremental Generation of Formal Concepts in FCA

Abstract:The theory of Formal Concept Analysis provides efficient methods for conceptualization of a problem domain. The incremental construction method is used for problems with dynamic changing contexts. In the presentation a novel cost approximation method and a novel optimized nested incremental concept set construction method are presented.

13:20-13:40 Imre Piller: Semantic Graph Representation of Agent Behaviors

Abstract: This work presents various agent behaviors represented by semantic graphs. The proper description of behavior has a key role in psychological and ethological researches, in robotics and in the design of multi-agent systems. The flexibility of semantic graphs makes available the analyzis and the comparison of the declarative and the procedural representations. It helps to recognize the common patterns among the considered behaviors. The available evaluation and inference methods of the described behaviors have also mentioned.

13:40-14:00 Ahmed Bouzid: New Pose determination methods. A contribution to Positioning and velocimetry

Abstract: The estimation of the position (positioning) of a moving object remains a topical subject where many solutions are proposed on the literature. In the context of our researches, new methods are presented that are implemented within the framework of pose determination of ground vehicles. Starting from sensors conditioning, the process is based on hybrid reconfigurable technological targets which enhances the prototyping time. Some DSP (Digital Signal Processing) tools have been used and implemented on FPGA (Field Programmable Gate Array) as well as analog signal processing implemented on FPAA mainly as front end using a novel tool that we call MRVA (Multichannel Reconfigurable Voltage Attenuator). The presentation mainly focuses on the designed velocimeter and emphasis on the importance of a Jerkmeter for such applications. A GSS (Ground Speed Sensor) is implemented taking advantage of IF (instantaneous frequency) estimation and hybrid reconfigurability for handling (generating, preprocessing and processing) ultrasonic transducers signals for high velocity resolution.

14:00-14:20 Mohammad Alsaudi: Detecting Slow Port Scan Using Fuzzy Rule Interpolation

Abstract: Fuzzy Rule Interpolation (FRI) offers a convenient way for delivering rule based decisions on continuous universes avoiding the burden of binary decisions. In contrast with the classical fuzzy systems, FRI decision is also performing well on partially complete rule bases serving the methodologies having incremental rule base creation structure. These features make the FRI methods to be perfect candidate for detecting and preventing different types of attacks in an Intrusion Detection System (IDS) application. This presentation aims to introduce a detection approach for slow port scan attacks by adapting the FRI reasoning method. A controlled test-bed environment was also designed and implemented for the purpose of this study. The proposed detection approach was tested and evaluated using different observations. Experimental analysis on a real test-bed environment provides useful insights about the effectiveness of the proposed detection approach. These insights include information regarding the detection approach's efficacy in detecting the port scan attack and in determining its level of severity. In the discussion the efficacy of the proposed detection approach is compared to the SNORT IDS. The results of the comparison showed that the SNORT IDS was unable to detect the slow and very slow port scan attacks whereas the proposed FRI rule based detection approach was able to detect the attacks and generate comprehensive results to further analyze the attack's severity.

 

Széchenyi 2020

Partnereink