Emrullah Fatih Yetkin

Yetkin, Emrullah Fatih

Name Variants

Y., Emrullah Fatih
Yetkin, E. F.
Emrullah Fatih, Yetkin
Yetkin, Emrullah Fatih
Yetkin, E.
Fatih Yetkin E.
Yetkin, EMRULLAH FATIH
E. Yetkin
YETKIN, Emrullah Fatih
Emrullah Fatih Yetkin
E. F. Yetkin
YETKIN, EMRULLAH FATIH
Yetkin,E.F.
Yetkin E.
Yetkin,Emrullah Fatih
Emrullah Fatih YETKIN
Y.,Emrullah Fatih
EMRULLAH FATIH YETKIN

Job Title

Dr. Öğr. Üyesi

Email Address

fatih.yetkin@khas.edu.tr

ORCID ID

0000-0003-1115-4454

Full item page

Scholarly Output

8

Articles

3

Citation Count

8

Supervised Theses

3 Scholarly Output Search Results

Now showing 1 - 8 of 8

Citation Count: 3
ON SOFT ERRORS IN THE CONJUGATE GRADIENT METHOD: SENSITIVITY AND ROBUST NUMERICAL DETECTION
(SIAM PUBLICATIONS, 2020) Yetkin, Emrullah Fatih; Cools, Siegfried; Yetkin, Emrullah Fatih; Giraud, Luc; Schenkels, Nick; Vanroose, Wim
The conjugate gradient (CG) method is the most widely used iterative scheme for the solution of large sparse systems of linear equations when the matrix is symmetric positive definite. Although more than 60 years old, it is still a serious candidate for extreme-scale computations on large computing platforms. On the technological side, the continuous shrinking of transistor geometry and the increasing complexity of these devices affect dramatically their sensitivity to natural radiation and thus diminish their reliability. One of the most common effects produced by natural radiation is the single event upset which consists in a bit-flip in a memory cell producing unexpected results at the application level. Consequently, future extreme-scale computing facilities will be more prone to errors of any kind, including bit-flips, during their calculations. These numerical and technological observations are the main motivations for this work, where we first investigate through extensive numerical experiments the sensitivity of CG to bit-flips in its main computationally intensive kernels, namely the matrix-vector product and the preconditioner application. We further propose numerical criteria to detect the occurrence of such soft errors and assess their robustness through extensive numerical experiments.
Citation Count: 0
Assesment of soft error sensitivity of power flow analysis
(Gazi Univ, Fac Engineering Architecture, 2023) Yetkin, Emrullah Fatih
Today's power systems are large and interconnected to each other with many buses, lines, loads, and generators. Even the solution of a single snapshot of the system for specific conditions requires the solution of systems of equations with large sizes. Thus, to obtain the results in a reasonable time for large problems like electrical power flow simulations, modern large computational environments should be employed. However, because of the increasing number of components in the modern computational environment, the possibility of soft errors also increases. Soft errors can be defined as failures arising from several fluctuations due to x-rays, cosmic particle effects, etc. These types of errors usually appear at any time of computation as a bit-flip in any floating-point operations. In this paper, we will investigate the soft-error effects on large-scale power flow simulations. Generally, power flow calculations are performed by using Newton Raphson Method. The system is modeled by nonlinear equations and the solution process requires a linear solver is employed to take the inverse of the Jacobian matrix at each iteration. In this study, the soft-error sensitivity of the numerical methods used in load flow was examined, and the problems that may be encountered were revealed.
Ölçeklenebilir manifold öğrenme kütüphanesi geliştirilmesi: Scaman
(2024) Yetkin, Emrullah Fatih; Yetkin, Emrullah Fatih
This thesis presents an exploration of manifold learning and dimensionality reduction techniques, which are crucial in the fields of data science and machine learning. The center of this study is the development and evaluation of 'Scaman (Scalable Manifold Library), a Python-based computational tool designed to implement these techniques. This thesis investigates the key manifold learning algorithms. Including PCA,MDS, LE, and LLE and emphasizing the importance of eigenvalue solvers in these algorithms. The contribution of this thesis is the integration of advanced eigensolvers like NumPy, SLEPc and FEAST into key manifold algorithms within scaman package. The empirical analysis was conducted using various synthetic and real-world datasets. Those analyses focused on the efficiency, accuracy, and practical utility of scaman in different scenarios. Results demonstrate the tool's effectiveness, especially in handling large datasets. The advantages of FLANN and SLEPc prove scaman's efficiency in the creation of adjacency matrices and eigenvalue computation. The outcome of this thesis provides a computational tool for researchers and practitioners. Future directions include expanding the tool's capabilities by adding more algorithms, improving scalability, and applying various domain specific data-driven scenarios.
Recommendation of data visualization tools for non-technical people
(Kadir Has Üniversitesi, 2019) Yetkin, Emrullah Fatih; Yetkin, Emrullah Fatih
Big data analysis and data science are promising trends. Visualization is critical part. It outlines and presents data as information from different perspectives. Consequently, leaders, decision makers, and end users will grasp concepts and identify patterns with new dimensions. However, while time is still a complex dimension, the number of Information Visualization (InfoVis) software tools are increasing rapidly. This research test out how non-technical people select their InfoVis tools. Generally, end-users have factors affect the selection process of a software tool. A survey is used to detect these features and relations in between. Finally, results are checked and analyzes using python functions of visualization and machine learning functions to outline the grouping of features to simplify the selection process of software visualization tools. The outcome of this research can be used as a general guide to easier understand software visualization capabilities and to compare these tools from end users' perspectives. A framework will be introduced to categorize and suggest InfoVis tools to end users.
Makine öğrenmesinde endüstriyel veri mahremiyetinin üretken düşman ağları ve diferansiyel gizlilik kullanarak korunması
(2023) Yetkin, Emrullah Fatih; Yetkin, Emrullah Fatih
Yapay Zeka (AI) ve Makine Öğreniminin (ML) hızla yaygınlaşması, mahremiyetin korunmasına ilişkin endişeleri artırdı. Bu teknolojiler, endüstriyel IoT, sosyal medya ve çevrimiçi platformlar gibi kaynaklardan kişisel ve hassas bilgiler içeren ve gizlilik riskleri getiren kapsamlı veri kümelerine dayanır. Güçlü gizlilik koruma önlemlerinin alınması, AI ve ML uygulama risklerini azaltmak için çok önemlidir. Bu tez, AI ve ML sistemlerinde gizliliğin korunmasını incelemektedir. Araştırmamız, ML doğruluğunu korurken bir gizlilik koruma yöntemi geliştirmek için herkese açık veri kümelerinden yararlandı. Gizliliği artırmak için, yaklaşımımızı Diferansiyel Gizlilik (DP) ve Üretken Düşman Ağları (GAN) ile güçlendirdik. Etkinliğini altı gizlilik ölçüsü kullanarak değerlendirdik. Yaklaşımımız, ML performansından ödün vermeden gizliliği koruyarak fizibilite ve etkinlik göstermektedir. Ayıklanan özellik alt kümeleri, ML modelleriyle hassas verileri açığa çıkarabildiğinden, gizli hassas bilgilerin ortaya çıkarılması vurgulanmıştır. Yöntemin mahremiyet endişelerini ele almadaki etkinliğini deneysel bir çalışmada gösteriyoruz. Bulgular, AI ve ML sistemlerinde gizliliğin anlaşılmasına katkıda bulunur. Araştırma, bilgileri korumak için içgörüler ve yaklaşımlar sunarak güvenilir ML sonuçları sağlar. Bu çalışma, gizlilik bilgisini ilerleterek, gizliliğin korunmasında AI ve ML teknolojilerinin sorumlu gelişimini destekler.
Citation Count: 3
A spectral divide and conquer method based preconditioner design for power flow analysis
(2010) Yetkin, Emrullah Fatih; Dağ, Hasan
Power system simulations most of the time require solution of a large sparse linear system. Traditional methods such as LU decomposition based direct methods are not suitable for parallelization in general. Thus Krylov subspace based iterative methods (i.e. Conjugate Gradient Generalized Minimal Residuals (GMRES)) can be used as very good alternatives compared to direct methods. On the other hand Krylov based iterative solvers need a preconditioner to accelerate the convergence process. In this work we suggest a new preconditioner for GMRES which can be used in Newton iteration of power flow analysis. The new preconditioner employs the basic spectral divide and conquer methods and invariant subspaces for clustering the eigenvalues of the Jacobean matrix appears in Newton-Raphson steps of power flow simulation. ©2010 IEEE.
Citation Count: 0
A sparsity-preserving spectral preconditioner for power flow analysis
(TUBITAK Scientific & Technical Research Council Turkey, 2016) Yetkin, Emrullah Fatih; Dağ, Hasan
Due to the ever-increasing demand for more detailed and accurate power system simulations the dimensions of mathematical models increase. Although the traditional direct linear equation solvers based on LU factorization are robust they have limited scalability on the parallel platforms. On the other hand simulations of the power system events need to be performed at a reasonable time to assess the results of the unwanted events and to take the necessary remedial actions. Hence to obtain faster solutions for more detailed models parallel platforms should be used. To this end direct solvers can be replaced by Krylov subspace methods (conjugate gradient generalized minimal residuals etc.). Krylov subspace methods need some accelerators to achieve competitive performance. In this article a new preconditioner is proposed for Krylov subspace-based iterative methods. The proposed preconditioner is based on the spectral projectors. It is known that the computational complexity of the spectral projectors is quite high. Therefore we also suggest a new approximate computation technique for spectral projectors as appropriate eigenvalue-based accelerators for efficient computation of power flow problems. The convergence characteristics and sparsity structure of the preconditioners are compared to the well-known black-box preconditioners such as incomplete LU and the results are presented.
Citation Count: 2
Parallel implementation of iterative rational Krylov methods for model order reduction
(IEEE, 2009) Yetkin, Emrullah Fatih; Dağ, Hasan
Model order reduction (MOR) techniques are getting more important in large scale computational tasks like large scale electronic circuit simulations. In this paper we present some experimental work on multiprocessor systems for rational Krylov methods. These methods require huge memory and computational power especially in large scale simulations. Therefore these methods are fairly suitable for parallel computing. ©2009 IEEE.