Graduate School of Informatics

Graduate School of Informatics /Information Systems

In partial fulfillment of the requirements for the degree of Master of Science Fatih Ünver will defend his thesis.

Title: "USER MOBILITY BASED AUTHENTICATION AND CLOUD MOCKING" 

Date: 19th January 2018

Time: 14:30 AM

Place: A-108

Thesis Abstract :

Nowadays, mobile phones are widely used and carried by people almost all the time of a day. Also, people spend lots of time using these devices and applications inside. It means that these devices keep our personal data like application usage or environmental measurements that are produced and stored by build-in sensors on mobile devices even we do not use the device. This data can describe mobility of a human being easily. In this thesis, a cross-platform application that collects application usage statistics and sensory data and sends to the cloud periodically is developed. Then a virtual profile is created according to the data on cloud. Whenever user is trying to open an application, cloud looks previous mobility records of the user and decides whether this operation and environment is typical or not. This structure is applicable for multi-level authorization. Since the application needs to many cloud communication, we have also developed a cloud mocking framework. This framework can automatically create unit tests for all possible exceptions and success scenarios and these cases can be run and test without a real cloud connection. This framework is created to save time and effort in development and unit testing.

Graduate School of Informatics /Information Systems

In partial fulfillment of the requirements for the degree of Master of Science Fatih Kurt will defend his thesis.

Title: "ONTOLOGY GUIDED NEURAL NETWORK PROCESSING FOR AUTOMATIC IMAGE CAPTION GENERATION" 

Date: 24th January 2018

Time: 10:00 AM

Place: A-212

Thesis Abstract :

A fundamental problem in artificial intelligence is automatically describing the content of an image since it will help to identify objects with computer vision and create meaningful rich descriptions using natural language processing. At the same time, searching for an image with a description is another key element in proving theory work vice versa. In this work, ontology guided neural network processing will be used to further improve previous work regarding to image caption generation. By using ontology, more reliable results could be deducted from neural network processing by associating objects and their activities with vector results. Final trained system is expected to better process visual input and create better descriptions for images by also using RNNs(recurrent neural networks); thus, the work is aimed at further improving BLEU score of earlier works.

In result of interviews held for Cybersecurity Department MSc Non-Thesis Program for the 2017-2018 Spring Semester, the names of the students who are accepted to the programs are as follows:

For the accepted new students, there will be orientation meeting at February 07, 2018 between 11:00-12:00 in Class-02 of Informatics Institute. Courses, deficiency courses etc. will be explained in detail, therefore participation is mandatory.

In result of interviews held for Cognitive Science Department MSc and PhD Programs for the 2017-2018 Spring Semester, the names of the students who are accepted to the programs are as follows:

PhD

Kemal Aleçakır

Cumhur Çakmak

Giray Songül

Kadem Berker Yaşar

Master of Science

Sait Burak Akar

Ezgi Ateş

Fatma Betül Aydın

Burak Buruk

Şemsi Çakmak

Şeyma Nur Ertekin

Şura Genç

Emel Gülçek

Gizem Nur Özdemir

Aysun Saydam

Ali Can Serhan Yılmaz

Title: Neural Mechanisms Underlying Sub-Second Crossmodal Time Perception 

PhD Candidate: Utku Kaya

Program: Cognitive Science Department 

Date: 23 January Tuesday 15:30

Place: Conference Hall-01

Abstract: The integration of information from different senses is central to our perception of the world including the fundamental attributes of space and time. Audiovisual interactions have been particularly well studied in this context. Similar to the fact that visual stimuli can drive the perception of where a sound comes from (spatial ventriloquism), auditory stimuli can drive the perceived timing of visual events (temporal ventriloquism). These interactions are in accordance with the auditory system’s superior temporal and the visual system’s superior spatial resolution. The current study aims to understand the neural mechanisms underlying temporal ventriloquism. To achieve this objective, we collected behavioral and electroencephalography (EEG) data in tandem in four experiments. In Experiment-1, we have investigated how a single auditory event changes the perceived timing of a single visual event by varying the temporal offset between auditory and visual events. Experiment-2 focused on the mechanisms underlying the auditory influences on the visual time interval perception. In Experiment-3, we studied the effect of temporal ventriloquism on the perception of apparent motion by altering the timing of visual events leading to the motion percept. In the last experiment, we investigated the cortical mechanisms involved in the time interval adaptation aftereffects on the visual apparent motion. Overall, these experiments have significant contributions to our understanding of auditory influences on visual timing and cross-modal time perception in general. More generally, this study attempts to bridge the gap between our unified multisensory representation of the external world and low-frequency brain oscillations within the context of temporal ventriloquism.

In the result of Academic council decision for Software Management MSc Non-Thesis Program for the 2017-2018 Spring Semester, the names of the students who are accepted to the program are as follows:

Title: Optimization of Weights and Features in Use of AHP for SNP Prioritization

PhD Candidate: Arif Yılmaz

Program: Health Informatics Department Medical Informatics Program
Date: 16 January Tuesday 13:130

Place: Conference Hall-01

Abstract:

Single Nucleotide Polymorphisms (SNP) holds a promise in identification of genomic footprints of complex diseases such as cancer and diabetes diseases. However identification of SNPs associated to complex diseases is a challenging problem due to the high number and variety of SNPs present in individual genomes. Analysis of genome wide studies of SNP datasets mainly focus on statistical evidence.  As there are more than ten-millions of SNPs in human genome, incorporating biological and functional knowledge about statistically significant SNPs provides valuable features for further selection of SNPs. Analytical Hierarchy Process (AHP) based SNP prioritization approach is a method developed for this purpose. However, AHP requires expert knowledge, which results in subjective decisions. In this work, we propose a novel approach for AHP design and optimization by utilizing Random Forest based AHP (RF-AHP) assessment on categories. We utilize the results of previously developed genomic model on Prostate Cancer. Proposed RF-AHP approach was compared with Delphi-AHP based method on Schizophrenia, Prostate Cancer, Type 2 Diabetes and Alzheimer’s disease genomic datasets and same performance was achieved. Additionally, RegulomeDB database was integrated to RF-AHP. While similar performance was obtained in most of the datasets better prioritization scoring is achieved for Schizophrenia disease.

We would like to see you all in our GATEWAY 2018 event where our Multimedia Informatics Masters students will showcase their Game Development Pipeline course’s games and Machine Learning for Multimedia Informatics course’s projects they developed during this semester.

There will also be a League of Junior Games section where our young game developers will present their card and board games.
 
The event will be held at METU Informatics Institute on 17 January 2018 16:30-20:00. You are all kindly invited.

Do not forget to follow our Facebook page for more information: http://www.facebook.com/GATEWay.METU/
 

Graduate School of Informatics /Cognitive Science

In partial fulfillment of the requirements for the degree of Master of ScienceSavaş Çetin will defend his thesis.

Title: AUTOMATIC SENSE PREDICTION OF EXPLICIT DISCOURSE CONNECTIVES IN TURKISH WITH THE HELP OF CENTERING THEORY AND MORPHOSYNTACTIC FEATURES

Date: 02th February 2018

Time: 10:30 AM

Place: A-212

Thesis Abstract : Discourse connectives (and, but, however) are one of many means of keeping the discourse coherent. Discourse connectives are classified into groups based on their senses (expansion, contingency, etc.). They describe the semantic relationship of two discourse units. This study aims to build a machine learning system to predict the sense of explicit discourse connectives on the Turkish Discourse Bank data, which is manually gold-annotated. To do so, this study examines the effect of several features: i.e. transitions of Centering Theory and morphosyntactic characteristics of the arguments in a discourse relation. The results imply that Centering Theory, morphosyntactic features and their combinations affect each class of sense in a different way. When the base score is calculated with only the connective feature, the addition of Centering Theory features seems to have increased the predictions scores for Comparison and Expansion classes. Also, Tense, Aspect and Modality features are observed to slightly affect Temporal class in a positive way.

Due to the applications made for Cyber Security Department Non-Thesis MSc Program in 2017-2018 spring semester ,

Interviews will be held on 15 January 2018 Monday.

Invited applicants have to be ready in front of class II-01 (S-01) in Informatics Instute at the specified date