Araştırma Çıktıları | WoS | Scopus | TR-Dizin | PubMed
Permanent URI for this communityhttps://hdl.handle.net/20.500.14719/1741
Browse
10 results
Search Results
Publication Metadata only Combining spatial proximity and temporal continuity for learning invariant representations(2012) Kursun, Olcay; Aytekin, Tevfik; Kursun, Olcay, Department of Computer Engineering, Istanbul Üniversitesi, Istanbul, Turkey; Aytekin, Tevfik, Department of Computer Engineering, Bahçeşehir Üniversitesi, Istanbul, TurkeyLocation and time are two critical aspects of most security-related events, and thus, spatiotemporal data analysis plays a central role in many security-related applications. The human brain has great capabilities of developing invariant representations of objects by taking advantage of both spatial similarity of features of objects/events and their relative timings (temporal information). Trace learning rule is one well-known solution for this problem of combining temporal relations with spatial proximity in clustering tasks such as the one performed by self organizing maps. In this work, we investigate a two stage mechanism: i) finding local clusters using spatial proximity, ii) grouping these clusters as suggested by temporal continuity patterns. We show our experimental results on a movie created from face images. © 2012 IEEE. © 2013 Elsevier B.V., All rights reserved.Publication Metadata only How similar is rating similarity to content similarity?(2012) Başkaya, Osman; Aytekin, Tevfik; Başkaya, Osman, Department of Computer Engineering, Bahçeşehir Üniversitesi, Istanbul, Turkey; Aytekin, Tevfik, Department of Computer Engineering, Bahçeşehir Üniversitesi, Istanbul, TurkeyThe success of a recommendation algorithm is typically mea- sured by its ability to predict rating values of items. Al- Though accuracy in rating value prediction is an important property of a recommendation algorithm there are other properties of recommendation algorithms which are impor- Tant for user satisfaction. One such property is the diversity of recommendations. It has been recognized that being able to recommend a diverse set of items plays an important role in user satisfaction. One convenient approach for diversifi- cation is to use the rating patterns of items. However, in what sense the resulting lists will be diversified is not clear. In order to assess this we explore the relationship between rating similarity and content similarity of items. We discuss the experimental results and the possible implications of our findings. © 2014 Elsevier B.V., All rights reserved.Publication Metadata only Improving the performance of active voxel selection in the analysis of fMRI data using genetic algorithms(2013) Ülker, Ceyhun Can; Aytekin, Tevfik; Ülker, Ceyhun Can, Department of Computer Engineering, Bahçeşehir Üniversitesi, Istanbul, Turkey; Aytekin, Tevfik, Department of Computer Engineering, Bahçeşehir Üniversitesi, Istanbul, TurkeyRecent research has shown that it is possible to classify cognitive states of human subjects based on fMRI (functional magnetic resonance imaging) data. One of the obstacles in classifying fMRI data is the problem of high dimensionality. A single fMRI snapshot consists of thousands of voxels and since a single experiment contains many fMRI snapshots, the dimensionality of an fMRI data instance easily surpasses the order of tens of thousands. So, feature selection methods become a must from both classification and running time performance points of view. To this end several feature selection methods are studied, either general or specific to fMRI data. So far, one of the best such methods, which is specific to fMRI data, is called the active method [9]. In this work we combine genetic algorithms with the active method in order to improve the performance of feature selection. Specifically, we first reduce the feature dimension using the active method and search for informative features in that reduced space using genetic algorithms. We achieve better or similar levels of classification performance using a much smaller number of voxels than the active method offers. Copyright 2013 ACM. © 2014 Elsevier B.V., All rights reserved.Publication Metadata only Locality sensitive hashing based scalable collaborative filtering, Yerele Duyarli Kiyim Tabanli Ölçeklenebilir Işbirlikçi Filtreleme(Institute of Electrical and Electronics Engineers Inc., 2015) Aytekin, Ahmet Maruf; Aytekin, Tevfik; Aytekin, Ahmet Maruf, Bilgisayar Mühendisliʇi Bölümü, Bahçeşehir Üniversitesi, Istanbul, Turkey; Aytekin, Tevfik, Bilgisayar Mühendisliʇi Bölümü, Bahçeşehir Üniversitesi, Istanbul, TurkeyNeighborhood-based collaborative filtering methods are widely used in recommender systems because of their easy-to-implement and effective nature. One important drawback of these methods is that they do not scale well with increasing amounts of data. In this work we applied the locality sensitive hashing technique for solving the scalability problem of neighborhood-based collaborative filtering. We evaluate the effects of the parameters of locality sensitive hashing technique on the scalability and the accuracy of the developed recommender system. © 2021 Elsevier B.V., All rights reserved.Publication Metadata only An ensemble approach for multi-label classification of item click sequences(Association for Computing Machinery, Inc [email protected], 2015) Yağcı, A. Murat; Aytekin, Tevfik; Gürgen, Fïkret S.; Yağcı, A. Murat, Department of Computer Engineering, Boğaziçi Üniversitesi, Bebek, Turkey; Aytekin, Tevfik, Department of Computer Engineering, Bahçeşehir Üniversitesi, Istanbul, Turkey; Gürgen, Fïkret S., Department of Computer Engineering, Boğaziçi Üniversitesi, Bebek, TurkeyIn this paper, we describe our approach to RecSys 2015 chal-lenge problem. Given a dataset of item click sessions, the problem is to predict whether a session results in a purchase and which items are purchased if the answer is yes. We define a simpler analogous problem where given an item and its session, we try to predict the probability of purchase for the given item. For each session, the predictions result in a set of purchased items or often an empty set. We apply monthly time windows over the dataset. For each item in a session, we engineer features regarding the session, the item properties, and the time window. Then, a balanced random forest classifier is trained to perform pre-dictions on the test set. The dataset is particularly challenging due to privacy-preserving definition of a session, the class imbalance prob-lem, and the volume of data. We report our findings with re-spect to feature engineering, the choice of sampling schemes, and classifier ensembles. Experimental results together with benefits and shortcomings of the proposed approach are dis-cussed. The solution is efficient and practical in commodity computers. © 2017 Elsevier B.V., All rights reserved.Publication Metadata only Balanced random forest for imbalanced data streams, Dengesiz veri akimlari için dengelenmiş rassal orman(Institute of Electrical and Electronics Engineers Inc., 2016) Yağcı, A. Murat; Aytekin, Tevfik; Gürgen, Fïkret S.; Yağcı, A. Murat, Bilgisayar Mühendisliǧi Bölümü, Boğaziçi Üniversitesi, Bebek, Turkey; Aytekin, Tevfik, Bilgisayar Mühendisliǧi Bölümü, Bahçeşehir Üniversitesi, Istanbul, Turkey; Gürgen, Fïkret S., Bilgisayar Mühendisliǧi Bölümü, Boğaziçi Üniversitesi, Bebek, TurkeyData with highly imbalanced class distributions are common in real life. Machine learning application domains such as e-commerce, risk management, environmental, and health monitoring often suffer from class imbalance since the interesting case occurs rarely. Yet another layer of complexity is added when data arrives as massive streams. In such a setting, it is often of interest that a learning algorithm is updated in an incremental fashion for scalability and model adaptivity reasons while still handling the class imbalance. In this paper, we propose an ensemble algorithm for imbalanced data streams based on the offline balanced random forest idea. We also show on a recent dataset that the algorithm is useful for the buyer prediction problem in large-scale recommender systems. © 2017 Elsevier B.V., All rights reserved.Publication Metadata only Real time distributed analysis of MPLS network logs for anomaly detection(Institute of Electrical and Electronics Engineers Inc., 2016) Macit, Muhammet; Delibaş, Emrullah; Karanlik, Bahtiyar; Inal, Alperen; Aytekin, Tevfik; Badonnel, S.O.; Ulema, M.; Cavdar, C.; Granville, L.Z.; dos Santos, C.R.P.; Macit, Muhammet,; Delibaş, Emrullah,; Karanlik, Bahtiyar,; Inal, Alperen,; Aytekin, Tevfik, Department of Computer Engineering, Bahçeşehir Üniversitesi, Istanbul, TurkeyLarge scale IP networks contain thousands of network devices such as routers and switches. Massive amounts of logging data is generated by these devices. Analysing this data is both a challenge and an opportunity for finding network problems. Moreover, large IP networks contain devices from different vendors, so it is important to build a system which can work with network devices of different brands. In this study we describe a distributed architecture which can retrieve, store, and process massive amounts of network logging data in real time. Using this architecture we also build a basic anomaly detection system. The system statistically models cumulative counts of logs for different event types for all the devices in the network. The statistical approach lets the system to detect deviations from the normal behaviour without consulting expert knowledge. Our evaluations show that the system effectively handles massive amounts of data and detects anomalies. © 2016 Elsevier B.V., All rights reserved.Publication Metadata only On parallelizing SGD for pairwise learning to rank in collaborative filtering recommender systems(Association for Computing Machinery, Inc [email protected], 2017) Yaǧci, Murat; Aytekin, Tevfik; Gürgen, Fïkret S.; Yaǧci, Murat, Boğaziçi Üniversitesi, Bebek, Turkey; Aytekin, Tevfik, Bahçeşehir Üniversitesi, Istanbul, Turkey; Gürgen, Fïkret S., Boğaziçi Üniversitesi, Bebek, TurkeyLearning to rank with pairwise loss functions has been found useful in collaborative filtering recommender systems. At web scale, the optimization is often based on matrix factorization with stochastic gradient descent (SGD) which has a sequential nature. We investigate two different shared memory lock-free parallel SGD schemes based on block partitioning and no partitioning for use with pairwise loss functions. To speed up convergence to a solution, we extrapolate simple practical algorithms from their application to pointwise learning to rank. Experimental results show that the proposed algorithms are quite useful regarding their ranking ability and speedup patterns in comparison to their sequential counterpart. © 2017 Elsevier B.V., All rights reserved.Publication Metadata only Design and implemenatation of a job recommender system, Kariyer.net için İş İlan Öneri Sistemi Tasarmi ve Gerçekleştirimi(Institute of Electrical and Electronics Engineers Inc., 2017) Kara, Kemal Can; Esen, Samet; Kahyalar, Neşe; Karakaş, Aşkın Aşkn; Aytekin, Tevfik; Kara, Kemal Can, AR-GE Mühendisi, Istanbul, Turkey; Esen, Samet, Kariyer.net A.Ş., Istanbul, Turkey; Kahyalar, Neşe, Kariyer.net A.Ş., Istanbul, Turkey; Karakaş, Aşkın Aşkn, Kariyer.net A.Ş., Istanbul, Turkey; Aytekin, Tevfik, Bilgisayar Mühendisliǧi Bölümü, Bahçeşehir Üniversitesi, Istanbul, TurkeyRecommender systems help people to find items of interest by utilizing past user interactions (such as product views, ratings, and purchases). Today many e-commerce sites and large scale web applications use recommender systems and provide their customers personalized products. In this work we will share our recent experience in developing a job recommender system based on collaborative filtering at Kariyer.net. In particular, we will explain how and why we choose the recommender algorithm developed in the system, methods for evaluating success, and the system architecture. We will also mention future work that we plan to pursue for solving the problems we face in practice after this successful first attempt. © 2018 Elsevier B.V., All rights reserved.Publication Metadata only Short term water demand forecasting using regional data, Bölgesel veriler üzerinde yapilan kisa dönem su talep tahmini(Institute of Electrical and Electronics Engineers Inc., 2019) Zeynep Yildiz, Tugba; Aytekin, Tevfik; Zeynep Yildiz, Tugba,; Aytekin, Tevfik, Department of Computer Engineering, Bahçeşehir Üniversitesi, Istanbul, TurkeyLimited water resources and changing climatic conditions make water one of the critical natural resources. In order to manage this limited resource in the most effective way, real-time monitoring and automatic control systems are becoming increasingly popular. Water demand forecasting is one of the important subjects in these studies. Accurate water demand forecasting increases efficiency in the management of water networks and also allows for leak/fraud detection. In this work, we carry out short term water demand forecasting using water consumption data collected from water meters in a regional area. For forecasting, we first clean water consumption data, extract various features and apply machine learning methods for forecasting. After giving the experimental results we discuss future improvements. © 2020 Elsevier B.V., All rights reserved.