Prezentirani radovi na engleskom jeziku bit će poslani za uključenje u digitalnu bazu IEEE Xplore.

Program događaja

srijeda, 22.5.2019 15:00 - 19:00, Liburna, Hotel Admiral, Opatija Radovi  1. I. Stancin, A. Jovic (University of Zagreb Faculty of Electrical Engineering and Computing, Zagreb, Croatia) An Overview and Comparison of Free Python Libraries for Data Mining and Big Data Analysis  The popularity of Python is growing, especially in the field of data science. Consequently, there is an increasing number of free libraries available for usage. The aim of this review paper is to describe and compare the characteristics of different data mining and big data analysis libraries in Python. There is currently no paper dealing with the subject and describing pros and cons of all these libraries. Here we consider more than 20 libraries and separate them into six groups: core libraries, data preparation, data visualization, machine learning, deep learning and big data. Beside functionalities of a certain library, important factors for comparison are the number of contributors developing and maintaining the library and the size of the community. Bigger communities mean larger chances for easily finding solution to a certain problem. We currently recommend: pandas for data preparation; Matplotlib, seaborn or Plotly for data visualization; scikit-learn for machine leraning; TensorFlow, Keras and PyTorch for deep learning; and Hadoop Streaming and PySpark for big data. 2. F. Guzzi, L. De Bortoli, S. Marsi, S. Carrato, G. Ramponi (Università degli studi di Trieste, Trieste, Italy) Distillation of a CNN for a High Accuracy Mobile Face Recognition System  In face recognition systems, the use of convolutional neural networks (CNN) allows to achieve good accuracy performances, whose power derives extensively from a huge amount of well trained parameters. While using online services any mobile device can suffice for an accurate identification, in the offline scenario, implemented on a wearable mobile hardware, it is difficult to achieve both real-time responsiveness and high accuracy. In this paper we present a solution for replace an open source face recognizer network (provided as part of the dlib libraries), distilling its learned knowledge into a less demanding CNN, where the first one is used as an expert oracle that provides the targets, while the second is trained on the same input image, following a regression approach. In addition to lightness, our CNN is trained to use smaller faces by design, naturally allowing for recognition of identities in a wider distance range with a reduced amount of computation. This eventually permits the porting of the network to a dedicated mobile accelerating hardware. The hypothesis we want to demonstrate is that the feature space topology has been deeply explored during the training of the expert network, and due to the fact that no information is created during the upsampling of a tiny face to the input size of the expert oracle, the smaller network can provide the same accuracy at a reduced computational cost. 3. I. Hrga (Juraj Dobrila University of Pula/Faculty of Informatics, Pula, Croatia), M. Ivašić-Kos (Department of Informatics, University of Rijeka, Rijeka, Croatia) Deep Image Captioning: An Overview  Image captioning is a process of automatically describing an image with one or more sentence of a natural language. In recent years image captioning has witnessed rapid progress, from initial template-based models to existing, based on deep neural networks. This paper gives an overview of current issues and recent image captioning research, with a particular emphasis on models that use the deep encoder-decoder architecture. We discuss the advantages and disadvantages of different approaches, along with a review of some of the most commonly used evaluation metrics and datasets. 4. B. Malnar (Ericssson Nikola Tesla, Zagreb, Croatia), A. Unnervik, N. Kose Cihangir (Intel , Neubiberg, Germany) Using High Performance Computing in Vehicles to Create Image Datasets for Deep Learning   Modern vehicles are equipped with multiple sensors that help engineers to automate driving functions as much as possible. Among other types of sensors, these vehicles are equipped with multiple cameras that enable capturing the video data of the vehicles surroundings. These video data can be later used in training deep neural networks that are typically used within the vehicles to detect objects of interest around the vehicle. However, before we can use the video data for training, we have to annotate them. Nowadays, the annotation process is typically semi-automated, where the initial annotations are added in a datacenter and then fine-tuned by human annotators. In this paper, we investigate the possibility to utilize high performance computing inside the vehicles to already add initial annotations at the edge, instead of using the datacenter. With this approach, we expect a more scalable overall solution for the annotation process, because we utilize the computing power that is already available in the vehicles. For this investigation, we focus on Intel's heterogeneous automotive platforms, and analyze their effectiveness for the annotation tasks at the automotive edge. 5. N. Vrebčević, I. Mijić, D. Petrinović (Faculty of Electrical Engineering and Computing, University of Zagreb, Zagreb, Croatia) Emotion Classification Based on Convolutional Neural Network Using Speech Data  The human voice is the most frequently used mode of communication among people. It carries both linguistic and paralinguistic information. For an emotion classification task, it is important to process paralinguistic information because it describes the current affective state of a speaker. This affective information can be used for health care purposes, customer service enhancement and in the entertainment industry. Previous research in the field mostly relied on handcrafted features that are derived from speech signals and thus used for the construction of mainly statistical models. Today, by using new technologies, it is possible to design models that can both extract features and perform classification. This research explores the performance of a model that comprises a convolutional neural network for feature extraction and a deep neural network that performs emotion classification. The convolutional neural network consists of three convolutional layers that filter input spectrograms in time and frequency dimensions and two or more dense layers forming the deep part of the model. The unified neural network is trained and tested on utterances from the Berlin database of emotional speech transformed into spectrograms. Current preliminary results are encouraging. 6. L. Begic Fazlic (Umwelt-Campus Birkenfeld, Birkenfeld, Germany), A. Hallawa, A. Schmeink (RWTH Aachen, Aachen, Germany), A. Peine, L. Martin ( Uniklinik RWTH Aachen, Aachen, Germany), G. Dartmann (Umwelt Campus Birkenfeld, Birkenfeld, Germany) A Novel NLP-FUZZY System Prototype for Information Extraction from Medical Guidelines  Medical guidelines play an important role in the field of evidence-based medical treatment. The content of a medical guideline is based on a systematic review of clinical evidence with instructions and recommendations that clinicians can refer to. Most of the guidelines are available in unstructured text format. Hence, clinicians spend a significant amount of time to search and find relevant recommendations in their semantic context. Using Machine Learning algorithms, automatic information extraction from medical guidelines has recently become possible. In this work, we present a novel system for semantic extraction and the creation of a fuzzy rule database for clinical guidelines. The proposed system, dubbed as NLP-FUZZY, combines capabilities of Natural Language Processing (NLP) and Fuzzy Logic approaches. Firstly, NLP-FUZZY performs semantic extraction of medical guidelines using bi-directional LSTM, afterwards, using extracted semantic, it creates fuzzy rules, which is able to recognize new cases in a learning domain and predict the grade of recommendation. In order to test NLP-FUZZY, we compared its performance with state-of-the-art NLP approaches for clinical information extraction. Results show NLP-FUZZY superiority across the learning cycle: training, testing and verification. 7. G. Maltugueva, A. Yurin (ISDCT SB RAS, Irkutsk, Russian Federation) Improving Case-Based Reasoning with the Aid of Multi-Criteria and Group Decision-Making Methods   Solving semi-formalized and semi-structured tasks on the basis of a case-based reasoning (CBR) provides to use the accumulated knowledge and experience. Wherein, sometimes it is necessary to improve CBR, for example, to reduce the set of possible solutions and to ground their correctness not only with the use of a similarity measure. One way to solve this problem is to use mathematical multi-criteria and group decision-making methods. This paper discusses the combined use of the ARAMIS (Aggregation and Ranking Alternatives nearby the Multi-Attribute Ideal Situation), AIR (Aggregation of Individual Ranking), MOPC (Method Of Pairwise Comparison processing), COIP (Complex processing Of Individual Preferences) methods and a case-based reasoning. The algorithm and case study for solving problem of selecting construction materials in petrochemistry are described. The proposed approach provides grounding CBR results from the point of view of mathematical methods and growth their confidence from the point of view of the experts. 8. E. Cherkashin, V. Paramonov, A. Shigarov, A. Mikhailov (Institute of System Dynamics and Control Theory, Irkutsk, Russian Federation) Digital Archives Supporting Document Content Inference  The amount of documents created and circulated in the world increases day-to-day. Many of them are reused as a donor material for writing new documents. This often requires some time-consuming efforts on capturing, cleansing, and reformatting data that should be transferred from donor documents to recipient ones. The aim of the study is creating tools for developing digital archives to design and implement a domain-specific repositories of information objects. The paper presents a novel approach to the creation of digital archives aiming at an efficient data transfer among documents. The approach enables us to enrich documents by metadata, using Linking Open Data, documents templates, and logical inference. It was implemented based on contemporary open software tools. The implemented toolset was successfully used to generate some documents of educational process regulation in departments of Irkutsk State University. 9. V. Bogdanova, S. Gorsky (Matrosov Institute for System Dynamics and Control Theory of Siberian Branch of Russian Academy of , Irkutsk, Russian Federation) Multiagent Technology for Parallel Implementation of Boolean Constraint Method for Qualitative Analysis of Binary Dynamic Systems  The extensive use of binary dynamic systems (BDS) in scientific and applied research actualizes the developing of new methods for qualitative analysis of the BDS trajectories behavior. The high computational complexity of qualitative analysis problems requires the development of software and tools for its solution using parallel computing and service-oriented access to the resources of high-performance computing environments. We offer a new implementation technology of Boolean constraints method for the qualitative research of BDSs functioning on a finite time interval. According to this method, the satisfiability verification of dynamic system properties is reduced to solving the Boolean satisfiability problem using SAT and/or TQBF solvers. The implementation technology is based on models, methods, programs, and tools for automation of solving the quantitative analysis problems. Implemented as microservices, programs are installed on dedicated computational nodes. The agents of a semantic network, associated with installation nodes, run these microservices. Agents peer-to-peer interactions are based on a knowledge base of a subject domain. Agent behavior is based on a discrete event-driven model. We developed tools for constructing a dynamic property model in the form of Boolean constraint (a Boolean equation or a quantified Boolean formula) and parallel algorithms for solving the SAT and 2QBF problems. The results of computational experiments confirmed the effectiveness of the developed software based on the proposed technology. 10. K. Vidović (Ericsson Nikola Tesla, Zagreb, Croatia), S. Mandžuka, M. Šoštarić (Faculty of Transport and Traffic Sciences, Zagreb, Croatia) National Access Point for Provision of EU-wide Multimodal Travel Information Services   Multimodal travel information includes static or dynamic travel and traffic data covering at least two modes of transport and allowing the possibility to compare transport modes. According to EU directives, EU-wide multimodal travel information services should be accurate and available across borders to various users, and they should be accessible through appropriate National access points. One of their main tasks is interaction with other EU national access points for the exchange of travel and traffic information, which EU member states should provide according to EU ITS Directive and some Delegated regulations. This paper provides a proposal for deployment of Nation access point for multimodal travel information services in Republic of Croatia. A brief analysis of regulations that define requirements, key stakeholders and required data categories are presented in paper. Also, system architecture and functional requirements are presented on conceptual level. 11. P. Georgieva, R. Dolchinkov (Burgas Free University, Burgas, Bulgaria) Fuzzy Models for Managing a Micro Grid PV System   In this paper, models of two fuzzy inference systems for supporting the process of managing micro grid PV systems are presented. The first model is for managing a PV/battery grid-connected system and uses fuzzy mapping of three input variables: power produced by PV panels, power in battery and consumed power. The output of the FIS, obtained after applying the fuzzy rules, is a decision for choosing which power source is to be used. The second model consists of one additional energy source (diesel generator/ fuel cell) and thus the input variables are 4. The simulation tests are conducted in MatLab. This research is part of a project for optimizing the energy consumption of a building with the use of independent alternative renewable energy sources. The input test data for the optimization are the amount of used energy for lighting, heating, computers power supply and other needs of one particular building (the building of Burgas Free University) and the corresponding price of each of the used energy sources. 12. I. Barić (HEP d.d., Osijek, Croatia), R. Grbić, E. Nyarko (Fakultet Elektrotehnike, računarstva i informacijskih tehnologija Osijek , Osijek, Croatia) Short-Term Forecasting of Electricity Consumption Using Artificial Neural Networks - an Overview   The short-term forecasting of electricity consumption has emerged as an important area of research with the aim of increasing the efficiency and reliability of energy system operation. It plays a very important role in the field of scheduling, load analysis, planning and maintenance of the power system. Artificial Neural Networks (ANNs) are computational techniques which have been successfully applied in a number of different problems, including short-term electricity consumption forecasting. This paper gives an overview of the recently published research papers that deal with short-term electricity consumption forecast based on ANNs. The related papers are evaluated with respect to several aspects, such as applied ANN type, used input variables and obtained precision regarding short-term electricity consumption prediction. četvrtak, 23.5.2019 9:30 - 13:00, Liburna, Hotel Admiral, Opatija Radovi  1. G. Li, P. Rezaee Borj, L. Bergeron, P. Bours (Norwegian University Of Science and Technology, Gjøvik, Norway) Exploring Keystroke Dynamics and Stylometry Features for Gender Prediction on Chatting Data  Online anonymity is considered as one of the great gifts of the Internet, but it also brings dangers to society, such as cybercrime, online sexual abuse and bullying, and love scams. Many people are fond of chatting online to make new friends, but how can they be sure that the person sitting behind the other computer is really the person they claim to be? By studying stylometry and keystroke dynamics features from chat data, it is feasible to reveal the actual gender of an online user. In this paper, we examined stylometry and keystroke dynamics features from chat data, and proposed a Random Forest based gender prediction approach by analyzing these features. In order to evaluate the effectiveness of the proposed approach, a data acquisition was conducted to capture the keystroke dynamics and text information when participants were chatting remotely via Skype. All participants were invited to chat freely on any topic they preferred in order to get to know each other. Based on our experimental result, the proposed approach achieved 72% prediction accuracy by analyzing on this free-text data captured only in 15 minutes. 2. J. Grönman, P. Sillberg, P. Rantanen, M. Saari (Tampere University, Pori, Finland) People Counting in a Public Event—Use Case: Free-to-Ride Bus  In the case of traditional bus travel, there are many ways to gather the statistics of bus passengers. Paid tickets can directly contain information about on which stop the passenger came aboard and where he took off. Alternatively, stamping or reading travel cards can be used to figure similar statistics. In the case of free-to-ride bus service, these simple methods are generally not available. Yet, for further route and capacity planning producing exact – or at least informative – statistics is crucial. This paper presents a real-life use case of collecting statistics about bus passengers on a free-to-ride bus route on a large public event in the summer of 2018 in Pori, Finland. The use case utilized cost-effective and off-the-shelf components such as the Raspberry Pi 3 computer, position sensors, cameras and the Open Source Computer Vision Library version 3. The hardware and software components of the system, which was based on image analysis and shape detection, as well the results of the study, are explained in this paper. Furthermore, this paper presents a discussion on the challenges faced while developing and implementing the system. 3. D. Kinaneva, G. Hristov, J. Raychev, P. Zahariev (University of Ruse, Ruse, Bulgaria) Early Forest Fire Detection Using Drones and Artificial Intelligence  Abstract – Forest and urban fires have been and still are serious problem for many countries in the world. To fight forest fire, different solutions exists which aim to mitigate the damages caused by them using methods for early detection. In this paper we have discussed a new approach for fire control in which modern technologies are used. In particular we propose a platform that uses Unmanned Aerial Vehicles (UAVs) which constantly circle over potentially threatened by fire areas and also support the benefits of Artificial intelligence in terms of computer vision and recognition of smoke or fire of images or video which has been captured by drone’s cameras. Several different scenarios for the possible use of the drones for forest fire detection will be presented and analyzed, including a solution with the use of a combination between a fixed and rotary-wing UAVs. Keywords – early forest fire detection platform, drones, artificial intelligence, computer vision I. INTRODUCTION The most up to date information on the current fire season in Europe and in the Mediterranean area is provided by the European Forest Fire Information System EFFIS. Each year this institution provides annual report on the forest fires in Europe, the Middle East and North Africa. According to the last report they provided for 2017 the dramatic effects of wildfires have caused damages of over 1.2 million ha burnt natural lands in the EU and killed 127 people among fire fighters and civilians. Over 25% of the total burnt area was in the Natura 2000 network, destroying much on the efforts of the EU countries in preserving key biodiversity and natural habitats for future generations. The same report says that these fires caused estimated losses of around 10 billion euros. Despite this huge digits EFFIS also reports that there is a trend of decreasing the number of fires that have occurred during the last decade. This decrease can be explained with the more severe actions and sanctions towards the action of humans causing wildfires and with the introduction of more advanced technical solutions for early detection of fires. Obviously the fight against fires can mitigate the damages but the numbers that occurs in terms of burnt area territory and human lives are still huge. That’s why it is necessary and even a must to constantly develop, implement and upgrade systems for fire detection. The most important factors in the fight against forest fires include the earliest possible detection of a fire event, the proper categorization of the fire and fast response from the fire department. The aim of the platform that is to be proposed is not only to use modern technologies but on the contrary to improve the above mentioned factors by enhancing the fire detection, reducing the fake alarms and making emerging calls to fire services in case of detected no fake fire signal. In this paper a platform for early forest fire detection with two types of UAVs – fixed wings and rotary wings is proposed. Both drones will be equipped with cameras either optical, thermal or both. One of the drones – those with a fixed wings will permanently circle around the desired area and it will observe the territory. Since this drone flies at medium altitude (350 m to 5500 m) it can produce fake alarms because of its remote visibility. If fixed-wing UAV detects a fire this will trigger an alarm that would bring on the rotary-wing drone. That drone will go to the area where a fire is suspected via GPS coordinates provided by the medium altitude drone. Its role is to conform or neglect the alarm base on its observation. The rotary wing drone will fly to suspect area, will conform or neglect and it will go back to base station. It will not permanently circle around the area. The reason to use second drone is to mitigate the fake alarms. It will fly at low altitude (10 m to 350 m) and it will have better visibility. If the fire is confirmed another alarm is going to be triggered and the ground level teams are going to be informed. The platform is completely automated since both drones have on-board computing. They can detect fires based on data captured by the thermal cameras and they can process these data without the need for centralize computing engine. In addition to further improve the platform we have plan to implement artificial intelligence by allowing the drones to make fire predictions based on computer vision techniques. In order to implement artificial intelligence a neural networks are required. The neural networks are the hot topic in today’s computing systems because of their ability to “learn” how to perform tasks by considering examples, without being programmed or instructed to follow specific rules. The neural networks are inspired by the biological neural networks that constitute human brains. 4. P. Linna, N. Narra, J. Grönman (Tampere University, Pori, Finland) Intelligent Data Service for Farmers  The agricultural sector has been lagging in digital development. Development has long been based on increasing production by investing in bigger machines. Over the past decade, change has begun to take place in the direction of digitalization. One of the challenges is that different manufacturers try to get farmers data theirs own closed cloud-services. The worst, the farmers lose the overall views of theirs farms, because the data is located in different services. In this research is described project of MIKÄ DATA coals and general view to solve previous challenges. This project will build data intelligent data service for farmers, which is based on Oskari-platform. In the service farmers can see their own field data and many other data sources layer by layer. The project is focused on study machine learning techniques to develop yield prediction and find out correlation between many data sources. The peltodata-service would be ready end of the 2019. 5. D. Baeva (​​​​​​​​​​​​​"Angel Kanchev" University of Ruse, Ruse, Bulgaria), B. Baev (American University in Bulgaria, Blagoevgrad , Bulgaria) Semantic Approach in Encoding of the Meanings in Bulgarian Folklore Embroidery in Digital Libraries  The purpose of this article is to describe the architectural approach in building a knowledge base for the Bulgarian national embroidery, in which its subjective characteristics are thoroughly presented. There is an opportunity, along with the generally accepted features, to store the characteristics encoded as a negative or hidden element and recognized by historians. Ontology-based on subjective semantics would be extremely useful. We list a range of potential uses and offer types of measurements and applications that can be performed to measure the capabilities of the semantic approach. 6. A. Stipić, T. Bronzin, B. Prole (CITUS, Zagreb, Croatia), K. Pap (University of Zagreb/Faculty of Graphic Arts, Zagreb, Croatia) Deep Learning Advancements: Closing the Gap  This article explains how recent development in the field of Artificial Intelligence (AI) makes gap between human and machine smaller than ever before, by explaining and comparing traditional approach user in development of AI systems with new approach that has been used by AI system AlphaZero, developed by DeepMind. Traditionally AI systems have been tested in chess and the same has been done to demonstrate the power of AlphaZero. But, instead of playing against human, it played against the best (at the time) chess program Stockfish. While chess programs (before AlphaZero), were using powerful hardware and embedded built-in formal knowledge about the game, AlphaZero is using completely new approach, running on standard hardware and using deep learning. It learned about the game by playing a large number of games with itself, learning in the process. Article will also explain what is so revolutionary in AlphaZero approach to AI and how this new approach can be used in different areas of processing visual information, bio-medicine, autonomous driving, robotics and AI generated images/videos of humans. 7. A. Stojanović (University of Applied Sciences, Zagreb, Croatia), N. Lazić (University of Zagreb, Faculty of Philosophy, Zagreb, Croatia) A Method for Estimating Variations in Speech Tempo from Recorded Speech  In this paper we describe a method for measuring variations in speech tempo from speech samples recorded from Croatian news channels, where the text of what was spoken is available. For speech recognition we use a feed-forward neural network trained with about 150 seconds of speech. To extract word boundaries, we created a speech-to-text aligner capable of finding an acceptable match between text and sequence of phonemes classified by the neural network. The aligner takes into consideration certain categories of phonemes for which the neural network has higher accuracy. Preliminary experiments show average alignment miss of about one to three phonemes, depending on the speaker, the content, and recording quality. 8. G. Mirceva, I. Ivanoska, A. Naumoski, A. Kulakov (Faculty of computer science and engineering, Ss. Cyril and Methodius University in Skopje, Skopje, Macedonia) Feature Selection for Improved Classification of Protein Structures   Many research groups work on analyzing the structures of protein molecules since that may help to gain knowledge that can be used for designing drugs. To understand the protein structures, it is very important to categorize them in corresponding classes. Therefore, protein classification is one of the main topics in bioinformatics. In this paper, we propose an approach for classifying protein structures. First, the characteristics of the proteins are extracted in corresponding feature vectors. Then, feature selection is made in order to reduce the dimensionality of the dataset, as well as to keep only the most significant features. For feature selection, we use various feature selection techniques. Finally, we build models by using different classification methods. The proposed approach is evaluated in details and also the benefits of applying feature selection are analyzed. 9. A. Naumoski, I. Ivanoska, G. Mirceva (Faculty of Computer Science and Engineering, Skopje, Macedonia) Analysing the Influence of Two Similarity Metrics on the Ant Colony Optimisation Based Fuzzy-rough Feature Selection Algorithm  As the amount of the data increases in volume, veracity, and value, and thus the number of the features rises, so the standard algorithms will find it difficult to process the data without help of huge computer power. However, the feature selection methodology offers help for this issue. In its core, the feature selection tries to find the most predictive input features for the output (target) feature. Feature selection combined with a hybrid variant of rough sets, fuzzy-rough sets provides fuzzy-rough feature selection that could offer better results in this task. To help fuzzy-rough methods to find optimal subsets, attempts have been made using feature selection mechanism based on Ant Colony Optimisation (ACO). This approach is applied to the problem of finding optimal subset of features in fuzzy-rough data reduction process by using different similarity metrics. In this paper, we investigate the influence of two fuzzy similarity metrics on the performance of the feature selection ACO strategy. The investigation is made by using several datasets. We experimentally compare several classical classification algorithms by using the AUC-ROC evaluation measure. Additionally, we show the benefits of making feature reduction. 10. I. Ivanoska, M. Milenkovski, S. Kalajdziski, K. Trivodaliev (Faculty of Computer Science and Engineering, Ss. Cyril and Methodius University, Skopje, Macedonia) Web Tool for Graph Embeddings Representation Techniques Evaluation  Graphs are important structures which are used in wide range of applications, therefore, many techniques for graph analysis are proposed. Graph embeddings are among the most popular techniques recently for graph analysis. The choice of the most appropriate graph embedding technique is a challenging task which depends on the graph type and the intended application. The aim of this paper is to present a web-based tool with a simple user interface that has been built which enables performance comparison of different models for transformation of a given graph of interest. The system allows graph embeddings models comparison for three different tasks – nodes classification, edges prediction and graphs reconstruction. Additionally, tool users can compare specific algorithms performance for different values of their hyperparameters. This accelerates the process of selecting the most appropriate model for graph transformation and enables the use of these models by a larger number of people in different scientific disciplines. četvrtak, 23.5.2019 15:00 - 19:00, Liburna, Hotel Admiral, Opatija Radovi  1. D. Marčetić (University of Zagreb, Faculty of Electrical Engineering and Computing, Zagreb, Croatia), L. Maleš (University of Split, Faculty of Humanities and Social Sciences, Split, Croatia), S. Ribarić (University of Zagreb, Faculty of Electrical Engineering and Computing, Zagreb, Croatia) Crowd Motion Pattern Detection at the Microscopic Level  In this paper we present a common-sense reasoning model for crowd motion pattern detection and behaviour analysis at the microscopic level. Information about the trajectories of individuals is represented with fuzzy predicates. The characteristic motion patterns and the behaviour of groups of individuals are described with fuzzy predicates and fuzzy functions, respectively. The usability of the proposed model is tested on a simulated crowd scenario. The obtained results show that the model supports the efficient representation of expert knowledge and detection of motion/behaviour patterns. The main contribution of this paper is a proposed model for motion pattern detection and classification based on fuzzy knowledge which can imitate common-sense reasoning. 2. I. Tomičić, P. Grd, M. Schatten (Fakultet organizacije i informatike, Varaždin, Croatia) Reverse Engineering of the MMORPG Client Protocol  Massively Multi-Player On-Line Role-Playing Games (MMORPG) represent a computer game genre where the human player is able to assume the role of a game character and control its actions in a well defined virtual world; the prefix “massively” denoting the vast number of players playing the game simultaneously. Cheating in such games is a fairly common occurrence and presents a major concern in network games as it is able to degrade the experience of “honest” players, with cheat prevention being considered as a subset of security issues for online games. In this paper we have presented a form of a network-level game cheating based on the MMORPG client reverse engineering method within the game Mana World, an MMORPG with client/server architecture. We have used the network protocol vulnerability of a non-encrypted data traffic in order to reverse-engineer the original data packets sent by the original game client, and replicated these packets in our own AI-based game client which enabled us complete control of the game character via Python code. 3. R. Čorić, M. Đumić, S. Jelić (Department of Mathematics, J. J. Strossmayer University of Osijek , Osijek, Croatia) A Clustering Model for Time-Series Forecasting   In this paper we consider a novel Integer Linear programming approach for the cluster-based piecewise linear fitting model used for time-series forecasting. There are several approaches in literature that aim to find a set of patterns which represent similar situations in the time series. In order to predict target variable, different types of fitting methods can be applied to set of data that belongs to the same pattern. We propose method that uses finding of patterns and piece-wise linear fitting in the same time, in order to minimize total absolute deviations between predicted and real values of target variable. We give experimental results about comparison of our method to some common approaches. 4. I. Tolovski, A. Kostovska, N. Simidjievski (Jozef Stefan Institute, Ljubljana, Slovenia), L. Todorovski (Faculty of Administration, Ljubljana, Slovenia), S. Džeroski, P. Panov (Jozef Stefan Institute, Ljubljana, Slovenia) Towards Reusable Process-Based Models of Dynamical Systems: A Case Study in the Domain of Aquatic Ecosystems  Storing metadata about models of dynamical systems in a machine readable form is one of the key steps towards their accessibility and reusability. In the domain of process-based modeling of dynamical systems, the task is to construct an explanatory model of a dynamical system from domain knowledge and data. In this paper, we present a workflow for annotation, storage and querying of process-based models specifically in the domain of aquatic ecosystems. To provide the vocabulary of key terms about the process-based modeling paradigm, we use the Ontology for Process-Based Modeling of Dynamical Systems (OntoPBM). Next, to capture the domain-specific characteristics, we extend OntoPBM with terms specific for aquatic ecosystems. The annotations for each process-based model are stored in an RDF triple store. This enables us to execute SPARQL queries on facts asserted in the annotations, as well as facts inferred from the domain knowledge encoded in the ontology. Finally, by following the proposed workflow, we generate the minimal information about a model, which takes us one step closer towards reusable research. 5. A. Ushakov, I. Vasilyev (Matrosov Institute for System Dynamics and Control Theory of SB RAS, Irkutsk, Russian Federation) A Parallel Heuristic for a k-Medoids Clustering Problem with Unfixed Number of Clusters  The k-medoids clustering model is a well-known NP-hard optimization problem which a number of widely-used clustering algorithms are based on. In the original k-medoids problem, the number of clusters k is a pre-specified parameter that must be defined before actual clustering. We address a modification of the k-medoids problem where the number of clusters is not pre-specified. Instead, it is a decision variable bounded by a monotonically increasing function of k added to the k-medoids objective function. We propose an effective hybrid heuristic algorithm for the modified k-medoids problem and its shared memory parallel implementation. Our approach is rested upon the approximation of dissimilarity matrix via nearest neighbors, dual search, and sophisticated variable fixing techniques. The main advantage of our algorithm is that it provides a dual bound for the objective value, thus allowing one to ascertain the optimality of a solution found. We present computational results on large-scale problem instances with millions of decision variables. 6. S. Kochemazov, A. Semenov (ISDCT SB RAS, Irkutsk, Russian Federation) Computational Study of Time Constrained Influence Maximization Problem under Deterministic Linear Threshold Model for Networks with Nonuniform Thresholds  The Influence Maximization Problem (IMP) consists in choosing a set of vertices in a network that maximizes the spread of influence under a specific influence model. It is one of the relevant problems in network science. In the present paper we consider the time constrained variant of this problem under the deterministic Linear Threshold (LT) model. Because of the deterministic nature of the influence model and additional time constraints the usually employed algorithms for solving IMP under nondeterministic LT model can not guarantee the quality of obtained solution. Thus we propose and study the algorithms for the problem in the considered formulation and compare their performance with the competition. 7. M. Krišto, M. Ivašić-Kos (Department of Informatics, University of Rijeka, Rijeka, Croatia) Thermal Imaging Dataset for Person Detection  In this paper will be presented an original thermal dataset designed for training machine learning models for person detection. The dataset contains 7412 thermal images of humans captured in various scenarios while walking, running or sneaking. The recordings are captured in the LWIR segment of the electromagnetic (EM) in various weather condition- clear, fog and rain at different distances from the camera, different body positions (upright, hunched) and movement speeds (regular walking, running). In addition to the standard lens of the camera, we used a telephoto lens for video recording, and we compared the image quality at different weather conditions and at different distances in both cases and set parameters that provided the level of detail in the image that can be used to detect the person. 8. A. Shigarov, V. Khristyuk, A. Mikhailov, V. Paramonov (Matrosov Institute for System Dynamics and Control Theory of SB RAS, Irkutsk, Russian Federation) Software Development for Rule-Based Spreadsheet Data Extraction and Transformation  The paper proposes a knowledge-based software platform to generate applications for spreadsheet data extraction and transformation. The platform includes a flexible table object model and a domain-specific language for expressing user-defined rules of table analysis and interpretation. They serve to represent knowledge of table layout and content features, as well as their interpretation, depended on transformation goals. The platform enables translating such userdefined rules to Java programs. The generated source code is serialized as a project prepared for building an executable application by using the Maven tool. The execution of the generated application transforms spreadsheet data from arbitrary form defined by the rules to the canonical one. The empirical results demonstrate the applicability of the software platform to develop applications for converting data from arbitrary spreadsheet tables originated from various domains to relational flat file databases. 9. A. Berman, N. Dorodnykh, O. Nikolaychuk, A. Yurin (Matrosov Institute for System Dynamic and Control Theory of Siberian Branch of Russian Academy of Sc, Irkutsk, Russian Federation) Event Trees Transformation for Rule Bases Engineering  The paper describes an approach for rule bases engineering by transforming event tree diagrams which are applied in the field of failure and risk analysis of technical systems. The approach is based on the identification and extraction of structural cause-effect elements of event trees and their transformation into the elements of a target knowledge representation language, in particular, CLIPS. Description of an extended event tree formalism, an event tree metamodel and transformation technique are presented. An illustrative example describes the development of a rule-based knowledge base for diagnosing and forecasting the states of complex technical systems based on the approach proposed.

Slobodan Ribarić (Croatia), Andrea Budin (Croatia)

Popust se ne odnosi na studente doktorskog studija.

Slobodan Ribarić
Fakultet elektrotehnike i računarstva
Unska 3
10000 Zagreb, Hrvatska

Tel.:+385 1 612 99 52
Fax: +385 1 612 96 53
E-mail: slobodan.ribaric@fer.hr

Andrea Budin
Ericsson Nikola Tesla d.d.
Krapinska 45
10000 Zagreb, Hrvatska

Tel.:+385 1 365 34 23
Fax: +385 1 365 3548
E-mail: andrea.budin@ericsson.com 

Najbolji radovi bit će nagrađeni.
Prihvaćeni radovi bit će objavljeni u zborniku radova s ISSN brojem. Prezentirani radovi na engleskom jeziku bit će poslani za uključenje u digitalnu bazu IEEE Xplore.
.............
Postoji mogućnost da se odabrani znanstveni radovi uz određenu doradu objave u međunarodnom časopisu Journal of Computing and Information Technology (CIT).

Opatija, sa 170 godina dugom turističkom tradicijom, vodeće je ljetovalište na istočnoj strani Jadrana i jedno od najpoznatijih na Mediteranu. Ovaj grad aristokratske arhitekture i stila već 170 godina privlači svjetski poznate umjetnike, političare, kraljeve, znanstvenike, sportaše, ali i poslovne ljude, bankare, menadžere i sve kojima Opatija nudi svoje brojne sadržaje. 

Opatija svojim gostima nudi brojne komforne hotele, odlične restorane, zabavne sadržaje, umjetničke festivale, vrhunske koncerte ozbiljne i zabavne glazbe, uređene plaže i brojne bazene i sve što je potrebno za ugodan boravak gostiju različitih afiniteta. 

U novije doba Opatija je jedan od najpoznatijih kongresnih gradova na Mediteranu, posebno prepoznatljiva po međunarodnim ICT skupovima MIPRO koji se u njoj održavaju od 1979. godine i koji redovito okupljaju preko tisuću sudionika iz četrdesetak zemalja. Ovi skupovi Opatiju promoviraju u nezaobilazan tehnološki, poslovni, obrazovni i znanstveni centar jugoistočne Europe i Europske unije općenito.

Detaljnije informacije se mogu potražiti na www.opatija.hr i www.visitopatija.com.