Boğaziçi Üniversitesi Bilgisayar Mühendisliği Bölümü'nden Haberlerhttp://www.cmpe.boun.edu.tr/index_tr.phpBachar Wehbi, Montimage

Network monitoring is a laborious challenging task that is vital for a network operator, a service provider or a corporate network infrastructure in order to keep the network operation stable, smooth and safe. Monitoring provides valuable real time and historical information to understand the network usage trends and dynamics. Monitoring has many applications in modern network systems including (but not limited to) performance and capacity monitoring, quality and user satisfaction monitoring, operation monitoring, security monitoring.

For years, monitoring consisted primarily of the collection of global traffic counters and indicators provided by the network elements through Simple Network Monitoring Protocol (SNMP). This approach is no longer sufficient. Many factors are setting forth the need for advanced network monitoring:

The evolving face of the Internet with the rise of traffic intensive popular applications (Youtube, Facebook, P2P, etc.) and the adoption of new technologies as Web2.0 are continuously changing the traffic mix and the usage trends.
The Mobile and wireless communications move towards broadband converged networks and applications combined with the proliferation of smart phones are bringing transcend requirements on connectivity and quality. Users now expect services to be delivered anywhere with a quality in par with fixed internet.
The vulnerabilities introduced by this “open world”: Critical infrastructures are more than ever open to the Internet, the dematerialization of corporate IT and the success of cloud services are pushing towards proactive mechanisms for detecting and preventing anomalies (due to attacks or misbehaviours).

In this context, Deep Packet Inspection (DPI) is considered as a catalyser in the shift towards advanced monitoring. DPI is the process of capturing network traffic, analyzing and inspecting it closely to determine accurately what is really happening in the network. This “core component” will feed the different monitoring applications with high added value information.



In this presentation, we will present network monitoring in general and cover in more details two monitoring applications namely: quality monitoring and security monitoring. Being a core component Deep Packet Inspection techniques and challenges will be discussed.

Tarih: 08.02.12 11:00
Yer: Dept of Computer Eng, AVS Seminar room (ETA 16)]]>Wed, 08 Feb 2012http://www.cmpe.boun.edu.tr/announcements/showWeeklyProgram_tr.php?eventid=222 Cagri Ozcaglar, RPI

Background: Strains of Mycobacterium tuberculosis complex (MTBC) can be classified into major lineages based on their genotype. Further subdivision of major lineages into sublineages requires multiple biomarkers along with methods to combine and analyze multiple sources of information in one unsupervised learning model. Typically, spacer oligonucleotide type (spoligotype) and mycobacterial interspersed repetitive units (MIRU) are used for TB genotyping and surveillance. Here, we examine the sublineage structure of MTBC strains with multiple biomarkers simultaneously, by employing a tensor clustering framework (TCF) on multiple-biomarker tensors.
Results: Simultaneous analysis of the spoligotype and MIRU type of strains using TCF on multiple-biomarker tensors leads to coherent sublineages of major lineages with clear and distinctive spoligotype and MIRU signatures. Comparison of tensor sublineages with SpolDB4 families either supports tensor sublineages, or suggests subdivision or merging of SpolDB4 families.
Conclusions: TCF on multiple-biomarker tensors achieves simultaneous analysis of multiple biomarkers and suggest a new putative sublineage structure for each major lineage. Analysis of multiple-biomarker tensors gives insight into the sublineage structure of MTBC at the genomic level.


Tarih: 05.01.12 14:15
Yer: Dept of Computer Eng, AVS Seminar room (ETA 16)]]>Thu, 05 Jan 2012http://www.cmpe.boun.edu.tr/announcements/showWeeklyProgram_tr.php?eventid=221 Okan İrfanoğlu / NIH Pediatric Neuroimaging, Diffusion Tensor MRI Center

Abstract:
Diffusion Tensor Imaging (DTI) is a relatively novel Magnetic Resonance Imaging (MRI) method for observing the macro-level properties of the diffusion of the water molecules, which enables us to analyze the structure of the human brain white matter. Since its invention in mid 1990s, DTI has been gaining tremendous popularity in the neuroscience community for its perceived ability to model the complex connectional architecture of the human brain. This talk will first provide the mathematical and physical foundations of diffusion tensor imaging. The second part will describe the typical and atypical applications of DTI and why great caution needs to be exercised for any DTI based analysis due to common misconceptions and limitations involved with this imaging modality.


Dr. Okan İrfanoğlu received his BS (2001) and MSc (2004) degrees from the Computer Engineering Department of Boğaziçi University. He obtained his PhD degree from the Computer Sciences and Engineering Department of Ohio State University, with a dissertation on diffusion tensor imaging. He is now a senior researcher at the NIH Institute at Washington DC, and conducts research within the Pediatric Neuroimaging Group, Diffusion Tensor Imaging MRI Center. His areas of expertise are computer vision, medical imaging, machine learning and computer graphics.


Tarih: 28.12.11 12:00
Yer: AVS Room]]>Wed, 28 Dec 2011http://www.cmpe.boun.edu.tr/announcements/showWeeklyProgram_tr.php?eventid=220 Hans D Mittelmann, School of Math&Stats, Arizona State University

As is well-known semidefinite relaxations of discrete optimization problems can
yield excellent bounds on their solutions. We present three examples from our
collaborative research. The first addresses the quadratic assignment problem and
a formulation is developed which yields the strongest lower bounds known for
larger dimensions. Utilizing the latest iterative SDP solver and ideas from
verified computing a realistic problem from communications is solved for
dimensions up to 512.
A strategy based on the Lovasz theta function is generalized to compute
upper bounds on the spherical kissing number utilizing SDP relaxations. Multiple
precision SDP solvers are needed and improvements on known results for all
kissing numbers in dimensions up to 23 are obtained. Generalizing ideas
of Lex Schrijver improved upper bounds for general binary codes are obtained
in many cases.

Bio:

Hans Mittelmann is a professor of Computational Mathematics at Arizona
State University. Prior to his appointment he was a professor at the
University of Dortmund. He has a PhD in Mathematics from the Technical
University of Darmstadt, where he also obtained the Habilitation. He has
written over 120 papers in Computational Mathematics and currently maintains
two of the most frequented websites in the area of optimization software.
His research has for more than 30 years been done in interdisciplinary
collaboration, lately nearly exclusively in optimization.
He is on the editorial board of several journals and book series including
Computational Management Science, Computational Optimization and
Applications, and International Series in Numerical Mathematics. Membership
in professional societies includes INFORMS, the Institute for Operations
Research and the Management Sciences and the Society for Industrial and
Applied Mathematics.



Tarih: 23.12.11 11:00
Yer: Dept of Computer Eng, AVS Seminar room (ETA 16)]]>Fri, 23 Dec 2011http://www.cmpe.boun.edu.tr/announcements/showWeeklyProgram_tr.php?eventid=218 Oya Aran

2008 doktora Mezunumuz Oya Aran, IDIAPta yürüttüğü projeleri anlatıyor.

Tarih: 14.10.11 16:00
Yer: CMPE TERAS]]>Fri, 14 Oct 2011http://www.cmpe.boun.edu.tr/announcements/showWeeklyProgram_tr.php?eventid=219 Can Kavaklıoğlu

Cmpe 579 seminar

In machine learning applications involving large multi dimensional data often tensors are used to represent the data. One of the most frequent operations is tensor factorization. The goal of factorization operations is to calculate more than one tensors which make up the given input tensor according to the given model. Factorization
benefits from parallelisation a lot due to its data driven
characteristics.


Tarih: 04.10.11 12:00
Yer: AVS Seminar room]]>Tue, 04 Oct 2011http://www.cmpe.boun.edu.tr/announcements/showWeeklyProgram_tr.php?eventid=217 Edip Toplan

MS Thesis Defense

Tarih: 29.09.11 10:00
Yer: AVS Seminar Room (ETA 16)]]>Thu, 29 Sep 2011http://www.cmpe.boun.edu.tr/announcements/showWeeklyProgram_tr.php?eventid=216 buradan ulaşabilirsiniz. Tüm öğrencilerimizin ellerine sağlık.



Ödül alan öğrenciler şöyle:



1 - Adem Efe Gencer ve Barış Bilgiç, Parallelizing Maxflow Problem Solver, Danışman: Can Özturan


2 - Volkan Cirik ve Murat Seyhan, Social Semantic Community, Danışman: Suzan Üsküdarlı


3 - Ozan İrsoy, Review Of Performance Measures And Statistical Tests To Compare Classification Algorithms In General And Bioinformatics Domains, Danışman: Ethem Alpaydın]]>Fri, 03 Feb 2012http://www.cmpe.boun.edu.tr/announcements/index_tr.php?id=1328273617 Tue, 31 Jan 2012http://www.cmpe.boun.edu.tr/announcements/index_tr.php?id=1328009387
"Olimpiyatlar’da aldığı madalyalarla üniversiteye sınavsız yerleşti" ]]>Tue, 09 Aug 2011http://www.cmpe.boun.edu.tr/announcements/index_tr.php?id=1312889976