Faculty of EngineeringComputer Engineering

Computer Engineering




Seminar Date: 13 September 2017 

Topic:  “Stabilizing Distributed Systems” 

Speaker: Prof. Dr. Hakan Karaata, Professor in the Department of Computer Engineering at Kuwait University



Seminar Date: 21 April 2017 

Topic:  Yazılım geliştirme yaşam döngüsü   

Speaker: Özgün Emre Uğurlu, NETAŞ 


Seminar Date: 21 April 2017 

Topic:  Redis, SignalR  teknolojileri uygulamaları   

Speaker: Ersel Bozkartal, NETAŞ



Seminar Date: 21 April 2017 

Topic:  Profesyonel iş hayatında bizi neler bekliyor?   

Speaker: Hakan Altınbaş, NETAŞ





Seminar Date: 14 April 2017 


Topic:  Yeni Sanayi Devrimi : Sanayinin Sayısal Dönüşümü   


Speaker: Alper Gerçek,  Kıdemli Tasarım Lideri, Teknoloji Transfer Müdürlüğü, Aselsan





Seminar Date: 7 April 2017 


Topic:  Türkiye IT pazarı ve Teknoloji Trendleri   


Speaker: Aslı Demir KoçkalResearch Manager, Systems & Infrastructure Solutions, IDC





Seminar Date: 7 April 2017 


Topic: Büyük Veri ve Analiz (Big Data & Analytics)


Speaker: Dr. Çağatay Talay,  Senior Data Architect - Big Data Platform at Digiturk




Seminar Date: 17 March 2017 


Topic: Teknoloji Dönüşümünde BT Güvenliğinin Yeri


Speaker: Yeşim Öztürk, Yazılım Pazarından Sorumlu Kıdemli Araştırma Analisti, IDC





Seminar Date: 10 March 2017 


Topic: Bilgi Güvenliği ve Kalite


Speaker: Hatice Ezgi Civelek, Türkiye Sınai ve Kalkınma Bankası Bilgi Güveliği ve Kalite Birimi





Seminar Date: 24 February 2017 


Topic: 5G Teknolojileri ve IoT, FPGA Programlama, Yapay Zeka Uygulamaları


Speaker: Ali TELLİ, Ar-Ge Bölümü Yöneticisi, HAVELSAN





Seminar Date: 17 February 2017 


Topic: Siber Güvenlik


Speaker: Alper Başaran, Garnizon Bilgi Güvenliği






Seminar Date: 26 Aralik 2016  

Topic:  Combining Two Different Techniques for Image Segmentation

Abstract: Clustering is the process of combining some set of data which show similar characteristics into one set, while seperating them among other parts of the data. Clustering is the main topic of data mining. Also it is a very crucial task for many other fields such as computer vision, image processing, pattern recognition, machine learning, etc. Clustering is called 'image segmentation' in image processing or computer vision. Image segmentation can be defined as the task of dividing image into different homogeneous regions containing several patches, whereas union of any two regions are non-homogeneous. Image segmentation plays a key role for many low-level computer vision applications such as object recognition, object tracking, medical imaging, face recognition and detection, etc. Since, homogenity criteria is application specific, image segmentation is a very hard job for low-level vision.  Existing image segmentation algorithms can be generally classified into three major categories, i.e., feature space based clustering, spatial segmentation, and graph-based approaches. Feature space based clustering approaches operate based on the color or texture, so that they ignore spatial information. There are parametric and non-parametric methods for feature space analysis. Parametric methods used in feature space analysis can lead to very wrong results due to parameters estimated by user. Parametric methods rely on training set for the parameters which must be well-tuned. Also, they are not suitable for arbitrary structured spaces. Sometimes, due to feature space overlap, discontinuities in the image are not preserved and irrelevant patches can be grouped together. So, a non-parametric, discontinuity preserving multi-mode seeking algorithm, so called mean shift, was proposed to find modes in the complex arbitrary structured density. However, for a good segmentation we must exploit spatial information also. Graph based approaches try to exploit both feature space and spatial information. In graph-based approaches pixels or regions are considered as vertexes and weight of edges are calculated based on both color or texture (feature space) and spatial distance. After constructing graph and its weight, the graph is partitioned into several homogeneous regions by minimizing some energy function. Normalized cut(Ncut) was the graph-based method considered in this study. Normalized cut is superior to other graph-based methods, however it is very source demanding. It is impractical to use Ncut for images whose size is bigger than 150x150. Hence, to overcome this deficit, in this work first, image is segmented coarsely by using mean shift and then Ncut was run on the regions found by mean shift. Running Ncut on regions rather than pixels reduced computation time drastically.    

Speaker: Abdulvahid Uçar



Seminar Date: 26 Aralik 2016  


Topic:  Power Characteristics and Modeling during Data Transmission over WiFi and Cellular Networks


Abstract: Energy consumption on sMarchphones is increasing rapidly with the increasing usage of applications that accessing the Internet over WiFi or cellular connection.  This causes the reduction of battery life of sMarchphones.  Current battery technology is not sufficient to meet the energy needs of networked applications for a long time. When the applications that need the network connectivity run for a long period, batteries in sMarchphones are draining before the end of the day. Therefore, the applications on sMarchphones should be made more energy efficient. To develop energy-efficient applications on sMarchphones, it must be known the factors affecting energy efficiency. There are some studies on this subject. In these studies, the energy consumption characteristics of data transmission over WiFi and cellular (HSDPA, LTE and so on) networks are investigated and some metrics affecting energy efficiency are identified, such as signal strength, traffic burstiness, network throughput and etc. Besides, performance and power characteristics of WiFi and the cellular networks are derived via measurements. Also, it is presented the power models of data transmission over WiFi and the cellular networks. In addition, an energy-aware scheduling algorithm is developed that computes optimal communication schedule for energy savings on the cellular networks.


Speaker: Kübra Uludağ



Seminar Date: 26 Aralik 2016  


Topic:  Incentive Mechanisms for User-Provided Networks


Abstract: By the great increase of usage sMarch mobile devices, internet connection is added to the list of commodities which people can share. In such user provided networks (UPNs), people generally share their internet connections with people who are known by the providers without profit making purpose. In order to make UPNs ubiquitous, there must be incentive mechanisms that encourage people to participate actively these kinds of networks and induce them to own crucial roles in the network which allows people to access to the internet provided by any user in UPN. In this presentation, I will present the roles and challenges of UPNs, how difficult to design such an incentive mechanism, proposed bargaining based scheme to set a fair medium between internet connection providers and consumers and virtual currency for satisfaction of participants.


Speaker: Kenan Cebeci



Seminar Date: 19 Aralik 2016  


Topic:  Opinion Mining


Abstract:  “Opinion Mining” is an interested area since online shopping sites and also microblogs have become widespread. Many people use this shopping sites (hepsiburada, amazon, ..) to buy what he/she wants. And also there are nearly hundreds of comments almost every product which helps to people to decide whether or not to buy the item, product or service. This customer reviews describe the experiences with the product/service are includes many important data like the product’s features. Feature selection is an important step in opinion mining (there are many steps in opinion mining and it is still growing). Customers express product opinions separately according to individual features of the product (for example: camera, ram, capacity can be features for the mobile phones). And try to find what customers say about of the features of a product or a service details about which aspects of a product people felt good or bad (or neutral).


Speaker: Muhammet Alkan



Seminar Date: 19 Aralik 2016 


Topic:  A Review on Predicting Student Graduation in Higher Education Using Data Mining


Abstract: Data Mining is increasingly used in higher institution to discover valuable information particularly in student databases. It brings to the concept of Educational Data Mining (EDM). EDM is a field that analyzes educational data to resolve its research issues using statistical, machine learning, and data mining algorithms. The emergence of EDM creates a good environment to all stakeholders in higher institution. One of them is graduation information. The data of graduated students from each academic year is an important source of information used as indicators of the school effectiveness by state and federal governments and accrediting agencies. Therefore, data mining can be used as an effective tool to predict graduation information of students to give insights to all stakeholders in higher institution whether the students are able to graduate on time, fast, late or even drop out from school. As a result, it assists the students on selecting suitable courses or field of study to their successful academic paths. In addition, it also assists the instructors to organize the curricula more effective to improve student academic performance. The purpose of this presentation is to provide an overview on data mining techniques that have been used to predict graduation information in higher institution through a survey of literature.


Speaker: Izhan  Fakhruzi



Seminar Date: 16 Aralik 2016


Topic: “TUBITAK, BİLGEM Tanitimi ve Projeleri” 


Speaker: Alper Ay, Birim Yoneticisi, TÜBİTAK



Seminar Date: 12 Aralık 2016


Topic :  Solving Graph Coloring Problem by Using Evolutionary Algorithms


Abstract: The graph coloring problem is one of the well-known optimization problems from the literature that tries to assign different colors to the vertices connected through an edge in a given graph. Its aim is to use minimum number of colors to color the vertices in the graph. This problem can be used to solve many practical and theoretical problems. The graph coloring problem is an NP-complete problem and genetic algorithms are widely used to solve the graph coloring problem. In this presentation, we will explore the proposed solutions for graph coloring problem. The papers try to solve two types of graphs. The first graph model is a weighted static graph where each node has a cost. The algorithms use novel based crossover and local search techniques to find a better solution for this graph. The second model is a dynamic graph where each node has a life time to stay alive with its edges in. At each time step, number of nodes and number of edges in the graph change according to relative parameters. Some standart crossover and mutation techniques described in the literature are applied to this graph and it is the first and recently published paper that tries to solve dynamic graph coloring problem.   


Speaker: Gizem Süngü



Seminar Date: 12 Aralık 2016


Topic:  Device To Devıce (D2D) Communıcatıon In Lte-Advanced Networks


Abstract: D2D communication in cellular networks provides devices directly communicating without base stations or another appliance. D2D communication technology is an important feature to LTE-Advanced systems. High performance, increased coverage, high data rate support, wireless peer-to-peer services are some of the advantages of this technology. With this advantages compared to available communication techniques, it has a wide range usage area such as network traffic offloading, public safety deployments, social networking, military applications, context-aware applications, multimedia applications. For example, a user may be informed about discounts while he/she in the supermarket, or she/he can share videos/photos with nearby friends using D2D in LTE-A technology. Many social networking applications need to discover and register users’ geographical locations, but this process typically made by user’s itself manually. This can be done automatically  using D2D Proximity Services. From a technical perspective, with communicating via D2D in LTE-A, UEs (user equipment or mobile users) gain high data rates, low battery consumption and less delays. The goal of my presentation is to give an overview about D2D communication in LTE-Advanced networks standards and specs which are defined in 3GPP Release 12, and continues working with Release 13, and analyze related  research works about device discovery mechanism in this area.


Speaker: Melike Sinem Uçum



Seminar Date: 9 Aralik 2016


Topic: “Fikri ve Mulki Haklar, Patent, Faydali Model” 


Speaker: Ersin Dereligil, Yonetim Kurulu Uyesi, DESTEK PATENT



Seminar Date: 2 Aralik 2016


Topic: "IGA Teknoloji ve Servisleri" 


Speaker: Ersin Inankul, IGA IT Direktoru, İstanbul Grand Airport



Seminar Date: 25 Kasim 2016


Topic: "Connected People - Connected World" 


Speakers: Kutsal Anıl, Genel Müdür, PAVOTEK

Seminar Date: 7 Kasım 2016

Topic: "Performance Analysis of Nature Inspired Heuristics for Survivable Virtual Topology Mapping"

Seminer Abstracti:  The high capacity of fibers used in optical networks, can be divided into many channels, using the WDM technology. Any damage to a fiber causes all the channels routed through this link to be broken, which may result in a serious amount of data loss. As a solution to this problem, the virtual layer can be mapped onto the physical topology, such that, a failure on any physical link does not disconnect the virtual topology. This is known as the survivable virtual topology mapping problem. In this study, we investigated the performance of two popular nature inspired heuristics, namely, evolutionary algorithms and ant colony optimization, in finding a survivable mapping of a given virtual topology while minimizing the resource usage. Our results show that both nature inspired heuristics perform remarkably well for this problem. Furthermore, both methods can obtain high quality solutions in less than a minute.


Speakers: Yrd.Doç.Dr. Fatma Corut Ergin, Marmara Üniversitesi, Bilgisayar Mühendisliği Bölümü, Öğretim Üyesi 



Seminar Date: 7 Kasım 2016


Topic: "Effects of UAV Mobility Patterns on Data Collection in Wireless Sensor Networks"


Seminer Abstracti:  Sensor nodes in a Wireless Sensor Network (WSN) can be dispersed over a remote sensing area e.g. the regions that cannot be accessed by human beings (inaccessible regions). In such kind of networks, data collection becomes one of the major issues. Getting connected to each sensor node and retrieving the information in time introduces new challenges. Mobile sink usage, especially the Unmanned Aerial Vehicle (UAV), is the most convenient approach to cover the area and access each sensor node in such a large scale WSN. However, the operation of the UAV depends on some parameters such as endurance time, altitude, speed, radio type in use, and the path. In this paper, we explore various mobility patterns of UAV that follow different paths to sweep the playground in order to seek the best area coverage with maximum number of covered nodes in less amount of time needed by the mobile sink. A realistic simulation environment is used in order to compare and evaluate the performance of the system. We present the performance results for the explored UAV mobility patterns. The results are very useful to present the tradeoff between maximizing the covered nodes and minimizing the operation time for choosing the appropriate mobility pattern.


Speakers: Yrd.Doç.Dr. Müjdat Soytürk, Marmara Üniversitesi, Bilgisayar Mühendisliği Bölümü, Öğretim Üyesi 



Seminar Date: 31 Ekim 2016


Topic: "Semantic Text Mining: Supervised and Semi-Supervised Classification Approaches"


Abstract:  Ever increasing volumes of structured and unstructured data in variety of types such text, multimedia, and graph is popularly referred as "Big Data".  An important portion of this data is in unstructured text format. There are important challenges to extract meaning or useful information from large textual datasets. One of the major methods used for organizing and mining big textual data is text classification. We use classification algorithms from machine learning domain which is a sub-field of Artificial Intelligence. In order for machine learning algorithms to work, unstructured textual documents are represented using Vector Space Models (VSM). Semantics of VSM comes from the distributional hypothesis, which states that words that occur in similar contexts usually have similar meanings. Although, the VSM  approach is very simple and commonly used, it has several limitations. One of them is the assumption of independency between terms as well as between documents. Documents are represented only with their term frequencies, disregarding their position in the document or their semantic or syntactic connections between other words. As a result, polysemous words (i.e. words with multiple meanings) are treated as a single entity and synonymous words are mapped into different components. This has a negative impact on performance of machine learning algorithms. There is a need for more advanced methods that incorporates additional semantics into these algorithms. In this seminar, we present several semantic text mining algorithms ranging from higher-order co-occurrence based probabilistic methods to class semantics based kernel methods for supervised and semi-supervised classification of textual documents.


Speakers: Yrd.Doç.Dr. Murat Can Ganiz, Marmara Üniversitesi, Bilgisayar Mühendisliği Bölümü, Öğretim Üyesi 



 Seminar Date: 31 Ekim 2016


Topic: "Migrating Birds Optimization: A New Meta-heuristic Approach and Its Application to the Quadratic Assignment Problem"


Topic:  We propose a new nature inspired metaheuristic approach based on the V flight formation of the migrating birds which is proven to be an effective formation in energy saving. Its performance is tested on quadratic assignment problem instances arising from a real life problem and very good results are obtained. The quality of the solutions we report are better than simulated annealing, tabu search, genetic algorithm, scatter search, particle swarm optimization, differential evolution and guided evolutionary simulated annealing approaches. The proposed method is also tested on a number of benchmark problems obtained from the QAPLIB and in most cases it was able to obtain the best known solutions. These results indicate that our new metaheuristic approach could be an important player in metaheuristic based optimization.


Speakers: Doç.Dr. Ali Fuat Alkaya, Marmara Üniversitesi, Bilgisayar Mühendisliği Bölümü, Öğretim Üyesi  



Seminar Date: 24 Ekim 2016


Topic: "Affective Computing: Challenges, Applications, Approaches"


Seminer Abstracti:  In human-human interactions we communicate using verbal messages and also non-verbal messages such as facial expressions, changes in our voice, and body gestures. The non-verbal messages reinforce or modify what is said in words. In the near future, it is envisioned that human-machine interaction will go beyond the mouse and the keyboard and will become more similar to human-human interaction. Machines will be aware of the emotional and mental state of humans and will adopt their responses accordingly. Affective computing is a relatively new research field the aim of which is to design systems that can recognize, interpret and simulate human emotions and related phenomena. In the first part of the seminar an introduction to the field of affective computing will be made, focusing on the application areas, and major research challenges.  In order to design systems which can recognize emotions, naturalistic training and test data are needed. Most databases available to researchers today are recorded in an acted way under laboratory conditions, and do not reflect real-world conditions. In the second part of the seminar, two new naturalistic affective audio-visual databases BAUM-1 and BAUM-2 will be introduced, which contain spontaneous Turkish expressions of 13 emotions and mental states. Then, three recent methods for multimodal audio/visual affect recognition will be described which use facial expressions and speech and fuse them at the decision level.


Speakers: Prof.Dr. Çiğdem Eroğlu Erdem, Marmara Üniversitesi, Bilgisayar Mühendisliği Bölümü, Öğretim Üyesi  



Seminar Date: 24 Ekim 2016


Topic: "Research Guide for Graduate Students"


Seminer Abstracti:  The aim of this presentation is to introduce strategies for graduate students to succeed in their study. The content is  based on both experience of the advisors on their graduate students and set of documents related with the topic. The research process will be covered by presenting a set of related issues  including  managing day-to-day work life of a graduate student,  communication with his/her advisor, reading technical papers, plagiarism, publishing results of research and professional development.   Unlike undergraduate courses,  graduate studies require a student read large volumes of technical documents.   As part of the presentation, critical reading strategies are briefly presented. Writing at the Master’s or PhD level is a challenging process, which will be covered briefly as the last phase  of the presentation


Speakers: Prof.Dr. Haluk Topçuoğlu, Marmara Üniversitesi, Bilgisayar Mühendisliği Bölümü, Bölüm Başkanı



 Seminar Date: 14 Ekim 2016


Topic: “CDN and  Cloud Technology”


Speakers: Atakan TİTİZ, DoraBase, IT Departman Direktörü



Seminar Date: 7 Ekim 2016


Topic: “IoT and Mobile Communication”


Speakers: Serhat Türkiçin, Turkcell, M2M ve IoT Servisleri ve Çözümleri Bölümü Yöneticisi. 



Seminar Date: 30 Eylül 2016


Topic: “CSE497 ve CSE498 Bitirme Projesi Hazırlama Süreçleri”


Speakers: Yrd. Doç. Dr. Müjdat Soytürk, Öğr. Gör. Birol Gençyılmaz



Seminar Date: 5 May 2016

Seminar Topic: "Vulnerability Analysis of Soft Errors"

Seminar Abstract:  As the technology advances, the transistors become smaller and the system becomes more vulnerable to the soft errors. Soft errors are transient errors that occur as bit flip in the memory and caused program's behaviour become unreliable. To prevent these errors some fault tolerance techniques are proposed. These techniques generally introduce information or hardware overhead to the system and increases the cost significantly. Consequently, selective reliability is proposed to reduce the overhead.  Some of the bit flips that occur at microarchitectural structures, don't affect the output of the program. The Architectural Vulnerability Factor (AVF) is the probability of a fault in structure affect the output. With this metric fault tolerance and recovery techniques may only apply most vulnerable parts and the overhead decreases. AVF provides error propagation probability of structures, however when it applies to the cache memories and register files, it cannot calculate probability accurately because of the characteristics of these structures. Hence, Cache Vulnerability Factor (CVF) and Register Vulnerability Factor are proposed to calculate the probability of error propagate to other components for cache and register respectively.  The vulnerability analysis also applies to the software level to quantify the vulnerability of a program.

Speaker: Zuhal Öztürk

Seminar Date: 5 May 2016

Seminar Topic: "Preserving High Performance in Cloud Computing with Optimal Power Consumption"

Seminar Abstract:  Cloud computing is one of the most studied subjects of all computing paradigms today. Even though researchers have intensive interest on this area, still there are many issues that have difficult challenges await to be solved. Power consumption of cloud computing systems is one of these issues. While managing energy consumption efficiently, cloud systems should not experience performance loss. In this presentation, IaaS(Infrastructure as a Service) -one of the core services of a cloud computing system- and virtualization will be introduced briefly. State-of-the-art techniques for energy saving in the cloud computing computing systems will be presented. Trade-offs of virtual machine consolidation between performance and power consumption will be discussed.

Speaker: Mustafa Gamsız

Seminar DAte: 29 April 2016

Seminar Topic: “Robotics and Artificial Intelligence”

Speaker: Nezih Ergin Özkucur, Boğaziçi Uni. Computer Eng. PhD student 

Seminar Date: 28 April 2016

Seminar Topic: "Fireworks Algorithm for Optimization Problems"

Seminar Abstract:  A novel swarm intelligence algorithm, called Fireworks Algorithm (FWA), is proposed for global optimization of complex problems. A comprehensive study on FWA revealed that the algorithm works surprisingly well on benchmark functions. The details of Fireworks Algorithm will be introduced. However, FWA has some limitations. A set of improvements applied on conventional FWA, named Enhanced FWA (EFWA) will be decribed. Also, three other variants of FWA, called Dynamic Search FWA (dynFWA) and Adaptive FWA (AFWA) and a novel GPU-based Fireworks Algorithm (GPU-FWA) will be presented and compared with conventional FWA.

Speaker: Hakan Pekdemir

Seminar Date: 28 April 2016

Seminar Topic: "MPI-Based Parallel Brain Simulations with NEURON"

Seminar Abstract:  Neurons (also known as nerve cells) are the cells that participate in the construction of the central and peripheral nervoussystem. Gathering information about neurons, helps us to understand causes and effects of the neurological diseases. Simulation of neurons and networks of can be used to gather information about neurons with the help of simulation tools such as NEURON. NEURON is a brain simulator that is widely used by neuroscientist. It is possible to simulate neurons and networks of them as if they exist by using NEURON. Unfortunately, simulation of neurons requires more computational resources as the number of neurons in simulation models increase and also requires more time to complete simulations. To overcome this problem, parallel simulations of models can be conducted. NEURON uses MPI (Message Passing Interface) library to provide communication between processors in parallel simulations. Neurons, NEURON simulator and MPI standard will be presented as parts of the seminar course presentation.

Speaker: Serap Korkmaz

Seminar Date: 21 April 2016

Seminar Topic: "Broadband Access Networks"

Seminar Abstract:  The evolution of the access network towards broadband is one of the most crucial and demanding challenges in the battle between the actors in the telecommunications industry. The issue of establishing access networks for broadband services is complex, both in terms of the broadband service market, and in terms of the future regulation of the customer access to telecommunications infrastructure. Moreover, there is a wide range of technologies available for broadband access, which further complicates the issue of broadband access. Broadband access technologies enable data, voice, video and other multimedia applications for home and business use. There are several rival technologies that are competing to deliver broadband connections each with its own distinct advantages and disadvantages. The choice of what access technologies to deploy depends mainly on its commercially viability and which access technology can best serve the current and future consumer demands.

Speaker: Merve Günaydın

Seminar Date: 21 April 2016

Seminar Topic"Convolutional Neural Networks"

Seminar Abstract: Convolutional Neural Networks (CNN) is one of the popular feedforward neural networks. CNN has demonstrated good performance on image classification in the large scale visual recognition challenge (ILSVRC2012) which is competition of ImageNet in 2012 and the goal of this competition is to estimate context of images. CNN’s success is the learning ability of the rich mid-level image representation. CNNs learn but they need huge amount of annotated image samples and they need millions of parameters for estimation. New generated GPUs have highly optimized implementation of 2D convolution and their performance is adequate for the training of huge CNNs.  Andrej Karpathy et. al. developed a deep neural network model for inferring hiding alignment between segments of sentences and the region of the image. They implemented a multimodal Recurrent Neural Network architecture that takes an input image and generates its description in text. They used CNN on the image side.

Speaker: Fırat Kurt

Seminar Date: 15 April 2016

Institution: NETAŞ Ar-Ge 

Seminar Topic: “eTicket Project, Business Intelligence, Data Warehousing, Database Management”

SpeakersOsman Kumaş (Software Design Manager– Netaş Promotion and eTicket Project), Oğuz Kutlu (Sr. Solution Architect– Database Management), Fatih Gündoğdu (BI Architect – Business Intelligence and Data Warehousing)

Seminar Date: 14 April 2016

Seminar Topic: "Reinforcement learning in nonstationary environments using spatiotemporal analysis"

Seminar Abstract: Traditional reinforcement learning (RL) approaches fail to learn a policy to attain a dynamic or nonstationary goal. The reason for this is that the RL agent cannot start learning the changed environment from scratch once it has converged to a policy of the environment before the change. While heuristic solutions where the RL agent is encouraged to use least  recently attempted actions are successful for slowly changing environments (Sutton & Barto, 1998). They they do not form a sufficiently fast solution to follow a nonstationary goal state that moves with the same velocity of the RL agent. The problem of finding an nearoptimal policy for a stationary environment is well dealt with by classic RL algorithms such as the TD learning algorithm (Sutton, 1988), SARSA algorithm (Rummery & Niranjan, 1994) and the Q learning algorithm (Barto et al.,1989). These algorithms generally model the environment by a goal state which upon the visit of the agent generates a positive reinforcement and by states that connect the start state to the goal state.While classical RL algorithms provide satisfactory solutions for the stationary environment problem very well, they fail in nonstationary environments. Some attempts to solve this problem have been made such as the DG learning algorithm (Kaelbling, 1993) where the values are not a function of stateaction pairs but of a triplet of (s,a,g) where the goal is considered to be dynamic. In this presentation we will discuss a new approach to problem where there is an adversarial relation present between the dynamic goal and the RL agent. To tackle this the spatiotemporal information of the dynamic goal state is incorporated, in terms of stochastic processes, as the rewards of the RL agent into the environment model thus enabling a modular solution to the problem.

Speaker: Burak Goncu

Seminar Date: 14 April 2016

Seminar Topic: "Coreference Resolution Sieve Based on Answer Set Programming"

Seminar Abstract: Coreference resolution is the task of determining linguistic expressions that refer to the same real-world entity in natural language. Formally, coreference consists of two linguistic expressions antecedent and anaphor. The anaphor is the expression whose interpretation (i.e., associating it with an either concrete or abstract real-world entity) depends on that of the other expression. The antecedent is the linguistic expression on which an anaphor depends. Coreference resolution is important for natural language understanding tasks like summarization, question answering, and information extraction. The Sieve  technique is a recently introduced and successful coreference resolution method. Sieve architecture applies a battery of deterministic coreference models one at a time from highest to lowest precision, where each model builds on the previous model’s cluster output.  Answer Set Programming is a general purpose logic programming formalism that supports comfortable representation of knowledge, nonmonotonic reasoning processes, and reasoning with hybrid knowledge bases. In an ASP logic program we describe (i) a set of potential solutions, (ii) relationships between concepts in the solution, and (iii) constraints on solutions. The building blocks for programs are atoms, literals, and rules.

We envision (i) to create a theoretical framework and a practical implementation for the Sieve: an assumption-based Sieve based on Answer Set Programming, we plan (ii) to perform extensions to the Sieve by using existing linguistic knowledge bases and other natural language processing resources, (iii) to extensively evaluate the new approach on international benchmark corpora for the English language. We envision (i) to create a theoretical framework and a practical implementation for the Sieve: an assumption-based Sieve based on Answer Set Programming, we plan (ii) to perform extensions to the Sieve by using existing linguistic knowledge bases and other natural language processing resources, (iii) to extensively evaluate the new approach on international benchmark corpora for the English language.

Speaker: Kenda Alakraa

Seminar Date: 8 April 2016

Institution: Huawei Ar-Ge

Seminar Topic: “Profiling User Interactions on Social Networks”

Speaker: Dr. Cüneyt Gürcan Akçora, Huawei Research Turkey, Technology Introduction Lab

Seminar Date: 24 March 2016

Seminar Topic: "Evolutionary Computation for Dynamic Optimization Problems"

Seminar Abstract: Dynamic optimization problems (DOPs) have been rapidly attracting the interest of the research community, since most real world problems in different domains have various characteristics of dynamism. One or more underlying elements of a given dynamic optimization problem including the objective function, problem constraints, decision variables or environmental parameters may change over time, where the main motivation becomes tracking the global optimum value as close as possible. In this study, the characteristics of DOPs are presented briefly  which is followed by the details of selected DOPs.  A set of selected evolutionary techniques for solving DOP will be presented as part of the talk. Finally, performance evaluation of selected techniques on a case study will be presented.

Speaker: Prof. Dr. Haluk Topcuoglu

Seminar Date: 24 March 2016

Seminar Topic: "Hierarchical Reinforcement Learning in Non-stationary Environments"

Seminar Abstract: A Reinforcement Learning (RL) agent mostly assumes environments are stationary which is not feasible on most real world problems. Most RL approaches adapt slow changes by forgetting the previous dynamics of the environment. A recent study on RL in non-stationary environments, Reinforcement Learning with Context Detection (RL-CD), presents a technique that helps determine changes of the environment's nature which the agent with the capability to learn different dynamics of the non-stationary environment. In this study we propose an autonomous agent that learns a dynamic environment by taking advantage of Hierarchical Reinforcement Learning (HRL) and present how the hierarchical structure can be integrated into RL-CD to speed up the convergence of a policy. We illustrate how our method outperforms RL-CD by employing hierarchies in RL.

Speaker:  Doç. Dr. Borahan Tümer

Seminar Date: 17 March 2016

Seminar Topic: "Improving Efficiency in Database Marketing"

Seminar Abstract: The companies, particularly the banking industry, regularly mount campaigns to improve customer value by offering new products to existing customers. In recent years, this approach has gained significant momentum because of the increasing availability of customer data and the improved analysis capabilities in data mining. Typically, response models based on historical data are used to estimate the probability of a customer purchasing an additional product and the expected return from that additional purchase. Even with these computational improvements and accurate models of customer behavior, the problem of efficiently using marketing resources to maximize the return on marketing investment is a challenge. Simply knowing a customer’s probability of responding to a particular offer is not enough when a company has several products to promote and other business constraints to consider in its marketing planning. This problem is compounded because of the capability to launch multiple campaigns through several distribution channels over multiple time periods. The combination of alternatives creates a complicated array of possible actions.

Speaker: Yrd. Doç. Mustafa Ağaoğlu

Seminar Date: 17 March 2016

Seminar Topic: "Auction And Barter Models For Electronic Markets"

Seminar Abstract: Recent advances in information technology have made major impact on markets and provided an ability to shift from the traditional physical markets where the traders meet at a certain place and at a certain time for exchanging commodities, to the electronic markets. However, in order for e-markets to be pervasive, innovative market mechanisms that support more complicated scenarios should be introduced. In this talk, two auction and barter based electronic market models will be introduced.

Our first model is a direct barter model for the course add/drop process in the universities. We model the course add/drop process as a direct barter problem in which add/drop requests can be placed as barter bids.

Our second model is the double auction with limited cover money model.

In this model, we propose the use of discrete time double auction institution for the trading of used goods as well as new ones. Our model allows declaration of an amount of cover money so that what is spent on purchased items minus the proceeds of sold items does not exceed this cover money amount. We also introduce a mechanism so that bidders may place multiple item requests in a single bid and limit the maximum number of items to be purchased.

Speaker:Yrd. Doç. Ali Haydar Özer

Seminar Date: 11 March 2016

Institutions: MİTTO(Marmara University's Innovation and Technology Transfer Office), ANONİMYA, REKMOB,Lojika Fieldlabs

Seminar Topic“Entrepreneurship Support Program and Sample Success Stories”

Speakers: Mustafa Tüysüz(MİTTO), Münevver Olcaysoy(Expert Advisor), Mustafa Burak Amasyalı(Lojika Fieldlabs), Taha Ali Çelik(ANONİMYA), Fatih Işık(REKMOB)

Seminar Date: 10 March 2016

Seminar Topic: "Modeling Variations First-Order Horn Abduction in Answer Set Programming using On-Demand Constraints and Flexible Value Invention "

Seminar Abstract: We study abduction in First Order Horn logic theories where all atoms can be abduced and we are looking for prefered solutions with respect to objective functions cardinality minimality, Coherence, or Weighted Abduction. We represent this reasoning problem in Answer Set Programming (ASP), in order to obtain a flexible framework for experimenting with global constraints andobjective functions, and to test the boundaries of what is possible with ASP, because realizing this problem in ASP is challenging as it requires value invention and equivalence between certain constants as the Unique Names Assumption does not hold in general. For permitting reasoning in cyclic theories, we formally describe fine-grained variations of limiting Skolemization. We evaluate ourencodings and extensions experimentally on the ACCEL benchmark for plan recognition in Natural Language Understanding. Our encodings are publicly available, modular, and our approach is more efficient than state-of-the-art solvers on the ACCEL benchmark. We identify term equivalence as a main instantiation bottleneck, and experiment with on-demand constraints that were used to elim- inate the same bottleneck in state-of-the-art solvers and make them applicable for larger datasets. Surprisingly, experiments show that this method is beneficial only for cardinality minimality with our ASP encodings.

Speaker: Yrd. Doç. Dr. Peter Schueller

Seminar Date: 10 March 2016

Seminar Topic: "Incentive Mechanisms in User Provided Networks "

Seminar Abstract:  Increase in the number of mobile users demanding internet connectivity gave rise to a cooperative solution called User-provided connectivity (UPC) which is a promising paradigm to remedy internet connectivity needs of mobile users. In this model, participants cooperate to enable a group of mobile users to access internet while they have no or limited connectivity. One of the key research problems in UPC is giving incentives to mobile network subscribers to promote active participation. In this talk, several incentive mechanisms that are proposed recently will be overviewed. Next, a new scheme will be described for incentivizing users in a network assisted User Provided Network (UPN) in which subscribers with limited or no connectivity access Internet while still charging to their own quota. This scheme proposes a simple and applicable bargaining based mechanism to encourage mobile network subscribers to join the protocol to provide UPC. Since the scheme is network assisted network operators are expected to support and do not impose any restrictions on the protocol. In this scheme utility and cost functions are defined for participants in terms of provided throughput and battery energy, and the protocol operates iteratively as in the case of a real world bargaining.

Speaker Yrd. Doç. Dr. Ömer Korçak

Seminar Date: 4 March 2016

Institution: ACM

Seminar Topic: “Ethics in Software Development”

Speaker: Lemi Orhan ERGİN

Seminar Date: 26 February 2016

Institution: SAP Türkiye 

Seminar Topic: “SAP and ERP Applications”

Speaker: İsmail Boz

Seminar Date: 19 February 2016

Seminar Topic: “CSE497 and CSE498 Thesis Completed Preparation Process”

Speakers: Yrd. Doç. Dr. Müjdat Soytürk, Öğr. Gör. Birol Gençyılmaz

Seminar Date: 4 December 2015

Institution:Migros Ar-Ge

Seminar Topic: "Retail and Technology"

Speaker: Kerim Tatlıcı, Information Technology & Business Development / Ar-Ge Center Director

Seminar Date: 27 November 2015

Institution: Türksat A.Ş.

Seminar Topic: “Türksat Applications ve Ar-Ge Activities”

Speakers: Prof.Dr. Ensar Gül,  (General Manager)   ve  Dr. Ahmet Savaş, (Manager, E-Devlet ve Bilgi Toplumu) 

Seminar Date: 20 November 2015

Institution: PAVOTEK

Seminar Topic: M2M and IoT Applications

Speaker: Kutsal Anıl, General Manager

Seminar Date: 23 October 2015

Seminar Topic: "DSL Fixed Access and Internet( İsmail GÜN), Databases and SQL (Mehmet Caner UZUNKONAK) "

Speakers: İsmail GÜN, Mehmet Caner UZUNKONAK

Seminar Date: 16 October 2015

Seminar Topic: "Build Automation with Gradle(Build automation is an essential method to package and deploy the software. It also helps us to see what is going on before sending bugs to customer's hands. Gradle is a build automation tool which takes advantages of previous build automation tools and introduces a Groovy-based domain-specific language.)"

Speaker: Çağlar Karabulut, Expert Software Engineer, ACM

Seminar Date: 9 October 2015

Seminar Topic: "Modeling “Reading between the Lines” Based on Scalable and Trainable Abduction and Large-scale Knowledge Acquisition"

Speaker: Asst. Prof. Dr. Naoya Inoue from Tohoku University, Japan. http://www.cl.ecei.tohoku.ac.jp/


Seminar Date: 8 May 2015

Seminar Topic: "Akademik Poster Hazırlama"

Speaker: Birol Gençyılmaz

Seminar Date: 6 May 2015

Seminar Topic: “Design Space Exploration for Customized Heterogeneous Multiprocessor System-on-Chip (MPSoC)”

Speaker: Prof. Smail Niar

Seminar Date: 17 April 2015

Institution: Koç-Sistem

Seminar Topic: “Bilgi Teknolojileri Servis Yönetimi”

Speaker: Dr. Turgay Karlıdere,  Koç Sistem Süreç, Hizmet ve Teknoloji Yönetimi – Grup Yöneticisi

Seminar Date: 10 April 2015

Institution: PAVOTEK

Seminar Topic: “FPGA Uygulamaları” (Gömülü Sistemler ve Robotik)

Speakers: Kutsal Anıl,  Genel Müdür

Seminar Date: 20 March 2015

Institution: Huawei Türkiye

Seminar Topic: “Backward Design and Plan”

Speaker: Y.Müh. Ali Uçar, Yönetici, Huawei Türkiye

Seminar Date: 13 March 2015

Institution: Huawei Türkiye Ar-Ge Merkezi

Seminar Topic: “Intelligent Search and Deep Neural Language Models”

Speaker: Araştırma Mühendisi Çağlar Tırkaz

Seminar Date: 6 March 2015 

Institution: ETİYA

Seminar Topic : M2M (Machine to Machine) Communication and IoT (Internet of Things)

Speakers: Dr. Ali Durmuş, ETİYA Genel Müdürü Rachmi Emir, ETİYA Iş Geliştirme Yöneticisi

Seminar Date: 12 December 2014

InstitutionLKD ( Linux Kullanıcıları Derneği)

Seminar Topic: "Özgür Yazılım ve Linux"

Speaker: Mahir B. Aşut

Seminar Date: 5 December 2014


Seminar Topic: "Etkili Sunum Teknikleri"

Speaker: Mehmet Ateş

Seminar Date: 28 November 2014

InstitutionsMİTTO(Marmara University's Innovation and Technology Transfer Office), KOSGEB(Küçük ve Orta Ölçekli İşletmeleri Geliştirme ve Destekleme Dairesi Başkanlığı), ANONİMYA, REKMOB

Seminar Topic: "Ar-Ge Inovasyon Destek Programları ve Desteklerden Yararlanan Mezunlarımızın Şirketlerinden Örnekler"

Speakers: Gözde Canlı (MİTTO), Emre Akın (KOSGEB), Ziya Mahmutoğlu (ANONİMYA), Fatih Işık (REKMOB)

Seminar Date: 21 November 2014


Seminar Topic: “Big Data, Application Areas and Technologies: An Overview”

Speaker: Abdülkerim Mızrak

Seminar Date: 7 November 2014


Seminar Topic: "Search Engine Optimization and Mobile Marketing "

Speakers: Burhanettin Mirzooğlu, Özkan Yılmaz

Seminar Date: 17 October 2014


Seminar Topic: "Guide for Evolving from Novices to Professional Software Developers"

Speaker: Lemi Orhan Ergin

Seminar Date: 10 October 2014

Seminar Topic: “Akademik Çalışmalarda Alıntı Yapma ve Meslek Etiği”

Speaker: Birol Gençyılmaz

This page updated by Computer Engineering on 18.09.2017 15:40:41