Araştırma Çıktıları | WoS | Scopus | TR-Dizin | PubMed
Permanent URI for this communityhttps://hdl.handle.net/20.500.14719/1741
Browse
31 results
Search Results
Publication Metadata only A threshold free clustering algorithm for robust unsupervised classification(2007) Temei, Turgay; Aydın, Nizamettin; Temei, Turgay, Department of Electrical Engineering, Fatih Üniversitesi, Istanbul, Turkey; Aydın, Nizamettin, Engineering Faculty, Bahçeşehir Üniversitesi, Istanbul, TurkeyA new information-theoretic, unsupervised, subtractive clustering algorithm is proposed. The algorithm eliminates threshold constraint to detect possible cluster members. Cluster centers are formed with minimum entropy. Instead of using a fixedthreshold, a decision region is formed with the use of maximum mutual information. Cluster members are chosen with a relative-cost assigned in partitions of data set. The algorithm yields more reliably distributed cluster numbers in statistical sense, hence reducing further computation for validation, which is justified for a set of synthetic data. © 2007 IEEE. © 2008 Elsevier B.V., All rights reserved.Publication Metadata only A novel information-theoretic clustering algorithm for robust, unsupervised classification(2007) Temel, Turgay; Aydın, Nizamettin; Temel, Turgay, Department of Electrical Engineering, Fatih Üniversitesi, Istanbul, Turkey; Aydın, Nizamettin, Engineering Faculty, Bahçeşehir Üniversitesi, Istanbul, TurkeyA new information-theoretic, unsupervised, subtractive clustering algorithm is proposed. The algorithm eliminates threshold constraint to detect possible cluster members. Cluster centers are formed with minimum entropy. Instead of using a fixed-threshold, a decision region is formed with the use of maximum mutual information. Cluster members are chosen with a relative-cost assigned in partitions of data set. The algorithm yields more reliably distributed cluster numbers in statistical sense, hence reducing further computation for validation, which is justified for a set of synthetic data. ©2007 IEEE. © 2008 Elsevier B.V., All rights reserved.Publication Metadata only Bearing fault detection in adjustable speed drives via a support vector machine with feature selection using a genetic algorithm(2008) Teotrakool, Kaptan; Devaney, Michael Joseph; Eren, Levent; Teotrakool, Kaptan, Department of Electrical & Computer Engineering, University of Missouri, Columbia, United States; Devaney, Michael Joseph, Department of Electrical & Computer Engineering, University of Missouri, Columbia, United States; Eren, Levent, Department of Electrical and Electronic Engineering, Bahçeşehir Üniversitesi, Istanbul, TurkeyThis paper presents a novel method to detect bearing defects in Adjustable Speed Drives (ASD's). The harmonics in pulse-width-modulation (PWM) input voltage waveforms and EMI noise in ASD systems make bearing fault detection more dfficult. The proposed method accomplishes bearing fault detection in ASD's by combining Motor Current Signature Analysis (MCSA), Wavelet Packet Decomposition (WPD), a Genetic Algorithm (GA), and a Support Vector Machine (SVM). The SVM in conjunction with the GA is applied to the rms values of the wavelet packet coefficients to obtain significant wavelet packet nodes which produce optimal classifiers for classifying both healthy and defective bearings in ASD systems. ©2008 IEEE. © 2008 Elsevier B.V., All rights reserved.Publication Metadata only Correspondenceless pose estimation from a single 2D image using classical mechanics(2008) Uǧurdaǧ, Hasan Fatih; Gören, Sezer; Canbay, Ferhat; Uǧurdaǧ, Hasan Fatih, Department of Electrical and Electronic Engineering, Bahçeşehir Üniversitesi, Istanbul, Turkey; Gören, Sezer, Department of Electrical and Electronic Engineering, Bahçeşehir Üniversitesi, Istanbul, Turkey, Department of Computer Engineering, Bahçeşehir Üniversitesi, Istanbul, Turkey; Canbay, Ferhat, Department of Electrical and Electronic Engineering, Bahçeşehir Üniversitesi, Istanbul, TurkeyEstimation of camera pose in world coordinates using a rigid object model with feature points and their respective images with known correspondence to the object model is a problem that has been studied extensively in the computer vision literature. We propose a correspondenceless method called Gravitational Pose Estimation (GPE). The word correspondenceless means the information of which image point corresponds to which object point is not needed. GPE requires only a single frame of image plus an object model and is inspired by classical mechanics. GPE creates a gravitational field from the image and lets the object model move and rotate in that force field. Experiments were carried out with both real and synthetic images. Results from real images were compared to measurements directly from our mechanical fixture. Experiments demonstrated GPE is accurate and hence allows efficient pose estimation even when correspondence information is unknown. © 2008 IEEE. © 2009 Elsevier B.V., All rights reserved.Publication Metadata only A generalization of Shannon's Mutual information for improved feature selection in databases involving possibly rare but well-predictable classes(2008) Kursun, Olcay; Kursun, Olcay, Department of Computer Engineering, Bahçeşehir Üniversitesi, Istanbul, TurkeyFeature selection is a critical step in many Artificial Intelligence and Pattern Recognition problems. Shannon's Mutual Information is a classical and widely used measure of dependence measure that can serve as a good feature selection algorithm. However, rare yet important regularities can be overlooked by this measure, which can create critical misses (false negatives) especially in the applications of biomedical fields, in which it is assumed that many factors contribute but in small amounts to the target function to learn. We propose a new measure of relevance which accounts for predictability of each signal from the other in the calculations which improves the feature detection capability. Also, this measure, in its formulation, turns out to be a generalization of Shannon's Mutual Information. © 2013 Elsevier B.V., All rights reserved.Publication Metadata only Augmenting clinical observations with visual features from longitudinal MRI data for improved dementia diagnosis(2010) Ünay, Devrim; Ünay, Devrim, Bahçeşehir Üniversitesi, Istanbul, TurkeyImage-based diagnosis in the medical area often requires qualitative interpretation from the experts, despite the high-resolution of the acquired images. Computation of quantitative measures and comparison of multiple patients using automated medical image analysis tools will help improve the diagnosis and efficiency, especially in the areas such as neurology, where diagnosis from one patient's data has limitations and the prevalence of neurodegenerative diseases is expected to substantially increase in the near future due to the aging population. To this end, this paper presents a novel work on fusing clinical and patient-demographics related observations with visual features computed from brain longitudinal MRI (magnetic resonance imaging) data for improved dementia diagnosis. Experiments with real data showed that augmenting cognitive scores with visual features from a subset of subcortical structures results in more accurate diagnosis. Moreover, subset of structures typically selected are consistent with those (being) investigated in the literature. Copyright 2010 ACM. © 2010 Elsevier B.V., All rights reserved.Publication Metadata only Determinig the ligand-specific regions of peptide-binding G-Protein Coupled Receptors(2010) Çobanoǧlu, Murat Can; Sezerman, Osman Uğur; Karabulut, Nermin Pinar; Çobanoǧlu, Murat Can, Faculty of Engineering and Natural Sciences, Sabancı Üniversitesi, Tuzla, Turkey; Sezerman, Osman Uğur, Faculty of Engineering and Natural Sciences, Sabancı Üniversitesi, Tuzla, Turkey; Karabulut, Nermin Pinar, Department of Computer Science, Bahçeşehir Üniversitesi, Istanbul, TurkeyThe G-Protein Coupled Receptor (GPCR) proteins are of paramount importance for pharmaceutical research as this single family of proteins is targeted by more than 50% of all modern drugs. Their ligands include but are not limited to small molecules, peptides and photons. Consequently it is vital to know the sites of interaction between the ligand and the receptor. These sites can be used for a multitude of purposes such as classification of sequences or drug design. We use the feature extraction method proposed in [1] to analyze the ligand-specific regions among various Peptide-binding GPCRs. We present the features that will help drug design researchers enhance specificity of peptide-binding drugs. © 2009 IEEE. © 2010 Elsevier B.V., All rights reserved.Publication Metadata only Enhanced feature selection from wavelet packet coefficients in fault diagnosis of induction motors with artificial neural networks(2010) Eren, Levent; Cekic, Yalcin; Devaney, Michael Joseph; Eren, Levent, College of Engineering, Bahçeşehir Üniversitesi, Istanbul, Turkey; Cekic, Yalcin, College of Engineering, Bahçeşehir Üniversitesi, Istanbul, Turkey; Devaney, Michael Joseph, College of Engineering, Columbia, United StatesWavelet packet decomposition (WPD) of line current has been successfully applied in motor fault detection. Enhanced feature selection from wavelet packet coefficients (WPCs) is presented in this paper. This method involves the decomposition of motor current into equally spaced frequency bands by using an all-pass implementation of elliptic IIR half-band filters in the filter bank structure to obtain WPCs in a computationally efficient way. Then, the bias in WPCs for each frequency band is removed to suppress both power system harmonics and leakage from adjacent frequency bands. Finally, the enhanced features are used as inputs to an ANN to provide motor fault detection with higher fault detection rate. © 2010 IEEE. © 2010 Elsevier B.V., All rights reserved.Publication Metadata only A hybrid method for feature selection based on mutual information and canonical correlation analysis(2010) Sakar, C. Okan; Kursun, Olcay; Sakar, C. Okan, Department of Computer Engineering, Bahçeşehir Üniversitesi, Istanbul, Turkey; Kursun, Olcay, Department of Computer Engineering, Istanbul Üniversitesi, Istanbul, TurkeyMutual Information (MI) is a classical and widely used dependence measure that generally can serve as a good feature selection algorithm. However, under-sampled classes or rare but certain relations are overlooked by this measure, which can result in missing relevant features that could be very predictive of variables of interest, such as certain phenotypes or disorders in biomedical research, rare but dangerous factors in ecology, intrusions in network systems, etc. On the other hand, Kernel Canonical Correlation Analysis (KCCA) is a nonlinear correlation measure effectively used to detect independence but its use for feature selection or ranking is limited due to the fact that its formulation is not intended to measure the amount of information (entropy) of the dependence. In this paper, we propose Predictive Mutual Information (PMI), a hybrid measure of relevance not only is based on MI but also accounts for predictability of signals from one another as in KCCA. We show that PMI has more improved feature detection capability than MI and KCCA, especially in catching suspicious coincidences that are rare but potentially important not only for subsequent experimental studies but also for building computational predictive models which is demonstrated on two toy datasets and a real intrusion detection system dataset. © 2010 IEEE. © 2010 Elsevier B.V., All rights reserved.Publication Open Access Use of line spectral frequencies for emotion recognition from speech(2010) Bozkurt, Elif; Erzin, Engin; Eroğlu Erdem, Çiğdem; Erdem, Tanju; Bozkurt, Elif, Koç Üniversitesi, Istanbul, Turkey; Erzin, Engin, Koç Üniversitesi, Istanbul, Turkey; Erdem, Cigdem Eroglu, Bahçeşehir Üniversitesi, Istanbul, Turkey; Erdem, Tanju Tanju, Özyeğin Üniversitesi, Istanbul, TurkeyWe propose the use of the line spectral frequency (LSF) features for emotion recognition from speech, which have not been been previously employed for emotion recognition to the best of our knowledge. Spectral features such as mel-scaled cepstral coefficients have already been successfully used for the parameterization of speech signals for emotion recognition. The LSF features also offer a spectral representation for speech, moreover they carry intrinsic information on the formant structure as well, which are related to the emotional state of the speaker [4]. We use the Gaussian mixture model (GMM) classifier architecture, that captures the static color of the spectral features. Experimental studies performed over the Berlin Emotional Speech Database and the FAU Aibo Emotion Corpus demonstrate that decision fusion configurations with LSF features bring a consistent improvement over the MFCC based emotion classification rates. © 2010 IEEE. © 2010 Elsevier B.V., All rights reserved.