Recent Advances in Precedence-Type Test and Applications

Ng, Hon Keung Tony

Department of Statistical Science Southern Methodist

University Dallas, Texas, U.S.A

Invited by: Bülent Karasözen

Place: IAM S212

Date/Time: 30.05.2017 -15.40

Abstract: In this talk, we will provide a comprehensive overview of theoretical and applied approaches to a variety of problems in which precedence-type test procedures can be used. Then, some recent advances on precedencetype tests, including the extension to progressively censored data, a two-stage test for stochastic ordering in two samples, and a sequential procedure for two-sample problem will be discussed. Collaborators: N. Balakrishnan (McMaster University, Canada), N. Kannan (National Science Foundation), M. L. Tang (Hang Seng Management College, Hong Kong), R. C. Tripathi (University of Texas at San Antonio), X. Zhu (McMaster University, Canada)

Short Bio : Professor Hon Keung Tony Ng

Hon Keung Tony Ng received the M.Sc. and Ph.D. degrees in statistics from McMaster University, Hamilton, ON, Canada, in 2000, and 2002, respectively. He is currently a Professor of Statistical Science with Southern Methodist University, Dallas, TX, USA. He is an Associate Editor of Communications in Statistics, Computational Statistics, Journal of Statistical Computation and Simulation, and Statistics and Probability Letters. His research interests include reliability, censoring methodology, ordered data analysis, nonparametric methods, and statistical inference. Dr. Ng is a fellow of the American Statistical Association, an elected senior member of IEEE and an elected member of the International Statistical Institute.

Generalization of leave-one-out cross-validation formula for linear statistical models

Savaş Dayanık

Industrial Engineering Department

Bilkent University

Invited by: Bülent Karasözen

Place: IAM S212

Date/Time: 23.05.2017 -15.40

Abstract: In comparing alternative linear models, one is interested in the mean squared error of each model on unseen data. This is often measured with cross-validation. However, as the number of folds increases, the computations can get prohibitively intense. It is well known that leave-one-out crossvalidation can be calculated by fitting each model to the entire data set only once. In this talk, we will show that the same is possible for any arbitrary k-fold crossvalidation.

Secure Data Outsourcing

Alptekin Küpçü

Department of Computer Engineering

Koç University

Invited by: Bülent Karasözen

Place: IAM S212

Date/Time: 16.05.2017 -15.40

Abstract: As many services are moving to the cloud, we are losing control of our own data. This constitutes one of the biggest problems that we face in this decade. In this talk, I will first describe various cloud scenarios and services, including cloud storage and databases, and searching within them. I will briefly present the problems and current solutions to them, highlighting techniques to know to understand and improve current solutions, as well as open problems.

Short Bio

Alptekin Küpçü received his Ph.D. degree from Brown University Computer Science Department in 2010. Since then, he has been working as an assistant professor at Koç University, and leading the Cryptography, Security & Privacy Research Group he has founded. His research mainly focuses on applied cryptography, and its intersection with cloud security, privacy, peer-to-peer networks, and game theory and mechanism design. He has also led the development of the Brownie Cashlib cryptographic library, which is available as open source online. Dr. Küpçü has various accomplishments including 3 patents granted, 7 funded research projects (for 5 of which he was the principal investigator), 2 European Union COST Action management committee memberships, 2 Koç University Teaching Innovation Awards, Science Academy Young Scientist Award (BAGEP), and the Royal Society of UK Newton Advanced Fellowship. For more information, visit https://crypto.ku.edu.tr

Tensor Decompositions in Machine Learning

Ali Taylan Cemgil

Department of Computer Engineering

Bogazici University

Invited by: Bülent Karasözen

Place: IAM S212

Date/Time: 09.05.2017 -15.40

Abstract: Matrix decompositions are widely used for developing models and expressing algorithms in signal processing, machine learning, data mining and related fields. This view, where the central object is a matrix has proven to be quite fruitful as it enables complex algorithms to be implemented using simple but powerful primitives, supported by a wide availability of tools for numerical computation. Yet, there are many situations when a matrix based description may become insufficient, or at best obscures the data model or the simplicity of an algorithm. In this talk, we will argue that multiway arrays with several indices, that we call tentatively as tensors, provide a natural framework for developing useful models for modern datasets as well as efficient algorithms for data processing. We express a tensor factorization models using a graph formalism reminiscent to probabilistic graphical models, reminiscent to tensor networks. The setting provides a structured and efficient approach that enables very easy development of application specific custom models, as well as algorithms for the so called coupled (collective) factorizations where an arbitrary set of tensors are factorized simultaneously with shared factors. Extensions to full Bayesian inference for model selection, via variational approximations or Monte Carlo methods are also feasible. We will also mention parallel and distributed inference algorithms and privacy preserving approaches to highlight more recent research directions. We will illustrate the approach in various applications.

Short Bio

Taylan Cemgil received Ph.D. (2004) from SNN, Radboud University Nijmegen, the Netherlands. Between 2004 and 2008 he worked as a postdoctoral researcher at Amsterdam University and the Signal Processing and Communications Lab., University of Cambridge, UK. He is currently an associate professor of Computer Engineering at Bogazici University, Istanbul, Turkey. His research interests are in Bayesian statistical methods and inference, machine learning and signal processing.

Reliability and Mean Residual Life Functions of Coherent Systems in an Active Redundancy

Konul Bayramoglu Kavlak

Department of Actuarial Sciences

Hacettepe University

Invited by: Bülent Karasözen

Place: IAM S212

Date/Time: 02.05.2017 -15.40

Abstract: In this talk the reliability and the mean residual life (MRL) functions of a system with active redundancies at the component and system levels are investigated. In active redundancy at the component level, the original and redundant components are working together and lifetime of the system is determined by the maximum of lifetime of the original components and their spares. In the active redundancy at the system level,the system has a spare, and the original and redundant systems work together. The lifetime of such a system is then the maximum of lifetimes of the system and its spare. The lifetimes of the original component and the spare are assumed to be dependent random variables. Key words. Coherent system, reliability function, active redundancy, bivariate order statistics, MRL functions.

Short Bio:

Konul Bayramoglu Kavlak took her B.Sc degree from Department of Statistics at Middle East Technical University, in 2006. Then, she attended the M.Sc program in Industrial Engineering Department at Bilkent University. She graduated from Bilkent University in 2009. She was awarded a grant by TÜBITAK in 2012 and she was a Research Fellow at The Ohio State University between 2012-2013. She received her doctoral degree in Statistics Department from Middle East Technical University in June 2014 . She was a Post Doctoral Visiting Scholar at the Department of Statistics at The Ohio State University between 2016-2017.  Currently, she is an Assistant Professor at the Department of Actuarial Sciences at Hacettepe University.  Her research interests focus mainly in statistics, probability, order statistics, reliability, and copulas. 

External Imbalances in Turkish Economy

Esma Gaygısız

Department of Economics

Middle East Technical University

Place: IAM S212

Date/Time: 18.04.2017 -15.40

Abstract:External imbalances have a crucial role in Turkish economy characterized by chronic current account and financial account deficits. The magnitudes of these crucial deficits, especially relative to the capability of the economy to cover these deficits, have had an increasing trend since 2002. The increasing dependency of the economy on external funds stands in stark contrast to its stagnant and even declining capability to pay back its obligations. This study investigates the compositions of Turkey’s external imbalances and relates these imbalances to the changing sectoral configuration and patterns of its production structures as well as its consumption directions.

Short Biography of the Presenter

Esma Gaygısız is an associate professor in the Department of Economics in Middle East Technical University (METU) in Ankara. She obtained her Doctor of Philosophy Degree in Economics and Econometrics from University of Manchester in United Kingdom. She received Master of Economics and Baccalaureate degrees from the Department of Economics, METU. Recently, she visited Norwegian University of Science and Technology between 2012-2014. In a crisis prone world, she concentrates on the links between real sectors and financial structures which constitute a vital dimension in understanding all economies in good and bad times. She studies on how real sector activities shape the financial outcomes and how distinct financial structures affect real sectors’ production characteristics and quantities.

Bounded Component Analysis: An Algorithmic Framework  for Blind Separation of Independent and Dependent Sources

Alper Erdoğan

Department of Electrical-Electronics Engineering 

Koç University

Invited by: Bülent Karasözen

Place: IAM S212

Date/Time: 11.04.2017 -15.40

Abstract: In many scientific experiments and engineering applications, measurements can be modeled as linear mixtures of  desired source signals, while the mixing system is unknown. The goal in Blind Source Separation (BSS) is to process these measurement sequences in an unsupervised manner to learn the inverse of the mixing  system, potentially by ​exploiting some side information about the sources. Most popular solution to BSS problem is Independent Component Analysis (ICA) which assumes that the sources are mutuallly statistically independent.  The fact that strong independence assumption may not hold in various scenarios has led to search for new frameworks that are capable of separating dependent sources.


Bounded Component Analysis (BCA) is a recent algorithmic BSS framework introduced in this direction. BCA can be considered as an extension of  the ICA framework  where the boundedness of sources is exploited to replace independence assumption with a weaker ``domain separability'' assumption. As a result BCA algorithms can be used to separate dependent as well as independent  signals from their linear mixtures.  In this talk, I'll introduce a geometric approach for developing instantaneous and convolutive BCA algorithms.  Furthermore, I'll illustrate the potential benefits of the corresponding BCA algorithms through different application examples.


This is a joint work with Huseyin A. Inan. and Eren Babatas

Short Bio

Alper T, Erdogan received his B.S.(93) in EE from Middle East Technical University in Turkey, M.S. (95) and Ph.D.(99) in EE from Stanford University. He worked as a principal research engineer in Globespan-Virata (formerly Excess Bandwidth) in Santa Clara CA during 1999-2001. In 2002, he joined Electrical-Electronics Engineering Department of Koc University in Istanbul, Turkey where he is currently an associate professor. Dr. Erdogan served as an associate editor for IEEE Transactions on Signal Processing and as a member of IEEE Signal Processing Theory and Methods Technical Committee. He is a recipient of TUBITAK Encouragement Award, Werner Von Siemens Award and Turkish Academy of Sciences Outstanding Young Scholar Award. His research interest is on Adaptive Signal Processing, Machine Learning, Communication Systems, Computational Neuroscience and Convex Optimization Applications.

A Statistical Approach in Turbine Heat Transfer

Harika Kahveci

Department of Aerospace Engineering


Invited by: Bülent Karasözen

Place: IAM S212

Date/Time: 07.03.2017 -15.40

Abstract: A high-quality extensive database is very critical to the gas turbine industry for improving the capabilities of the current state-of-the-art design of these machines so that more efficient cooled designs with extended turbine life can be accomplished. A series of experiments was performed at the OSU Gas Turbine Laboratory involving a rotating rig with a cooled 1-1/2 stage high-pressure transonic turbine operating at design corrected conditions with the goal of providing the turbine designer with such high-quality data. The turbine stage used is a modern 3-D design consisting of a cooled high-pressure vane, an uncooled high-pressure rotor, and a low-pressure vane. The work investigates the influence of different vane inlet temperature profiles and cooling flow rates on heat transfer of the full-stage turbine. A novel application of a traditional statistical method is introduced to the analysis to assign confidence limits to measurements in the absence of repeat runs. This approach is later incorporated into a CFD validation effort for blade heat transfer predictions in order to quantify the overall predictive uncertainty due to the variation in the inlet temperature profile, gauge position, and surface roughness. Presented data analysis highlights important turbine flow regions that are highly complex and are still not well understood today

Short bio:

Dr. Harika S. Kahveci is an Assistant Professor at Aerospace Engineering Department at METU. She holds a B.S. degree in Aeronautical Engineering from METU (2002), an M.Sc. degree in Aerospace Engineering from Penn State University (2004), and a Ph.D degree in Mechanical Engineering from The Ohio State University (2010). Before joining METU, she was employed at General Electric Company for eleven years where she held several positions in engineering and management.  She has expertise in hot gas path aero design and thermal design of gas turbine engines. During her Ph.D, Dr. Kahveci worked at the short-duration rotating rig facility of the Gas Turbine Lab at The Ohio State University investigating the film-cooled first-stage turbine vane heat transfer. In June 2015, Dr. Kahveci received the Gas Turbine Award, which is the prestigious ASME gas turbine industry award that recognizes outstanding contributions to the literature.

Enumeration of irreducible polynomials with prescribed coefficients

Emrah Sercan Yılmaz

College Dublin University

Place: IAM S212

Date/Time: 10.01.2017 -15.30

Abstract: In this seminar, we will give the general theory of enumeration of irreducible polynomials with prescribed coefficients from beginning. We will explain how these numbers are related with (fibre products of) supersingular curves.

A Parametric Simplex Algorithm for Linear Vector Optimization Problems

Firdevs Ulus

Department of Industrial Engineering
Bilkent University
Invited by: Murat Manguoğlu
Place: IAM-S209

Date / Time: 03.01.2017 / 15.40

Abstract. A parametric simplex algorithm for solving linear vector optimization problems (LVOPs) is presented. This algorithm can be seen s a variant of the multi-objective simplex algorithm. Different from it, the proposed algorithm works in the parameter space and does not aim to find the set of all efficient solutions. Instead, it finds a ‘solution’ which is a subset of efficient solutions that allows to generate the whole efficient frontier. In that sense, it can also be seen as a generalization of the parametric self-dual simplex algorithm, which originally is designed for solving single objective linear ptimization problems, and is modified to solve two objective bounded LVOPs with the positive orthant as the ordering cone. The algorithm proposed here works for any dimension, any solid pointed polyhedral ordering cone and for bounded as well as unbounded problems.Numerical results are provided to compare the proposed algorithm with an objective space based LVOP (Benson's) algorithm and with the multiobjective simplex (the Evans-Steuer) algorithm. The results show that for non-degenerate problems the proposed algorithm outperforms Benson's algorithm and is on par with the Evan-Steuer algorithm. For highly degenerate problems Benson's algorithm outperforms the simplex-type algorithms; however, the parametric simplex algorithm is computationally much more efficient than the Evans-Steuer algorithm for these problems.