Stories on Computer Engineering

Congratulations Dr. Goshgar Can Ismayilov!

Fri, 02 May 2025 00:00:00 +0000

Decentralized Privacy-Preserving Collective and Multi-Objective Trading Protocols on Blockchain with Zero-Knowledge Proofs

In this thesis, we propose decentralized privacy-preserving cryptographic protocols on blockchain with zero-knowledge proofs for three essential problems. Firstly, the privacy-preserving payment problem refers to a specific group of transactions where a source address (i.e. sender) transfers a certain amount of tokens to a destination address (i.e. receiver) while still protecting the privacy of their balances and transaction details. We extend this problem to address multi-token payments as well. Secondly, the privacy-preserving aggregation problem refers to a multi-party computation where a group of blockchain addresses (i.e. aggregators) aggregate their individual data to reach the global aggregation by still protecting the privacy of their own data. We also extend this problem to address prefix aggregation and to support for arbitrary numbers of aggregators. Thirdly, privacy-preserving multi-token bartering problem refers to a multi-party computation where a group of blockchain addresses (i.e. barterers) collectively exchanges a set of tokens in return for another set of tokens through proposing bids by still protecting the privacy of their balances and bids. We extend this problem to address multi-objective bartering using Bellman-Ford algorithm and pareto-domination. We propose the PTTS protocol for the first problem, the PVSS and PRFX protocols for the second problem and the PMTBS and zkMOBF protocols for the third problem. We analyze these protocols in terms of their scalability (as computational, communication and storage overheads) and their security (as potential attacks and reduction proofs). We perform experimental evaluation on Ethereum and Avalanche to measure blockchain gas consumption, proof generation/verification times and proof artifact sizes.

Congratulations Dr. Barış Yamansavaşçılar

Fri, 21 Mar 2025 00:00:00 +0000

Efficient Orchestration Methods in Air Computing Using Deep Reinforcement Learning

Abstract

After the proliferation of smart devices and diverse Internet of Things (IoT) requirements, we observe the dominance of cutting-edge applications with ever-increased user expectations in terms of mobility, pervasiveness, and real-time response. Over the years, to meet the requirements of those applications, cloud computing provides the necessary capacity for computation, while edge computing ensures low latency. However, these two essential solutions would be insufficient for next-generation applications due to the highly dynamic load. Therefore, we put forward a novel, next generation paradigm called air computing that presents a dynamic, responsive, and high-resolution 3D computation environment for all spectrum of applications. To this end, we first investigate efficient task orchestration through deep reinforcement learning (DRL) considering terrestrial resources via edge computing. Afterwards, to evaluate the corresponding air computing environments correctly, we design and develop our novel discrete event simulator, AirCompSim. Subsequently, we focus on the unknown user locations in an infrastructure-less environment in which users cannot connect to any communication device or computation-providing server, which is essential to task offloading in order to achieve the required quality of service. Thus, we propose a novel DRL-based scheme, DeepAir, which uses four main phases including sensing, localization, resource allocation, and multi-access edge computing. Performance evaluation results show that our DRL-based methods outperform the benchmark methods including heuristics and fuzzy logic. Thus, problems related to dynamic capacity enhancement and orchestration can be overcome in air computing environments.

A EuroHPC Success Story

Wed, 05 Mar 2025 00:00:00 +0000

Professor Akarun and her research team have been granted computing hours through EuroHPC, the European High-Performance Computing Joint Undertaking. In a recent interview, she discusses how access to supercomputers has propelled her research forward and explores the potential of Artificial Intelligence (AI) to make interactions with public services more seamless for the Deaf community.

In the interview, Professor Akarun explains:

Access to the EuroHPC infrastructure enables us to carry our research to a different level. Large language models have revolutionised the whole field of machine learning, but they are very costly to train and even fine-tune. With our previous supercomputing resources, training would take weeks or months, and we would exceed our allocated quotas before the models converged. Therefore, before the access now granted to us by EuroHPC, this kind of work would be out of the question.

Congratulations Dr. Yiğit Yıldırım

Fri, 10 Jan 2025 00:00:00 +0000

Abstract

The current raison d’être of robots is their potential to be utilized in tasks deemed suitable by humans and for the benefit of humanity. To ensure that robots perform designated tasks as successfully as humans, it is logical to aim for human intelligence or functionality. The research presented in this thesis is a collection of attempts to bring robot functionality one step closer to the human level. Learning from Demonstration (LfD), which essentially signifies teaching robots new skills by demonstrating them, provides invaluable resources for progressing in this direction. With the advances in data-driven approaches, LfD methods have undergone a significant transformation. Despite the improvements, there remains a substantial gap between humans and robots in terms of intelligence. To help bridge this gap, we first proposed increasing the sociability of mobile robots. We introduced a data-driven navigation framework that learns navigation-related movement primitives from real-world human data. The proposed model leverages spatial metrics of proxemics and trajectory characteristics to minimize disturbance to surrounding individuals. Building on this work, we proposed a new LfD framework called Conditional Neural Expert Processes (CNEP). CNEP learns movement primitives of diverse skills simultaneously in an unsupervised manner. Leveraging the entropy concept, CNEP paves the way for capturing the multimodalities in demonstrations. Finally, many real-world skills require the rhythmic application of the same primitive, such as hammering a nail. For such skills, we proposed the Position- Enhanced Movement Primitives (PEMP) model that learns to utilize the angular phase instead of the linear phase. Our experiments showed that PEMP has great potential in modeling and synthesizing periodic primitives.

Fatih Alagöz received the Excellence in Research Award

Sun, 01 Sep 2024 00:00:00 +0000

Boğaziçi University Foundation (BÜVAK) rewards scientists who make a difference with their research activities and publications at Boğaziçi University every year. This year, seven scientists in the senior and junior categories received the BÜVAK Academy Excellence Award. The Academic Board, which takes part in the determination of the scientists who receive the award, evaluates the national and international publication performances of the academicians. In this context, where different types of publications are published, their number, continuity and the number of citations they receive according to leading indexes are examined.

Congratulations Dr. Sümeyye Ağaç!

Tue, 27 Aug 2024 00:00:00 +0000

Enhancing Lightweight Models for Efficient Sensor-based Human Activity Recognition

Abstract

This thesis presents a comprehensive study on enhancing the performance of lightweight models for efficient sensor-based Human Activity Recognition (HAR) on resource-constrained devices, particularly in memory-constrained environments such as wearables with microcontrollers. By integrating convolutional block attention modules into the lightweight versions of the DeepConvLSTM model, originally proposed for sensor-based HAR, and the highly efficient SqueezeNet architecture, the study aims to boost recognition accuracy without increasing computational demands. The thesis examines the impact of channel and spatial attention mechanisms across various model sizes. Results demonstrate that attention-enhanced lightweight models can achieve performance levels comparable to larger, more resource-intensive models while maintaining minimal resource usage. Subsequently, combinations of knowledge distillation and attention mechanisms were applied to further improve model efficiency. It was found that attention-based distillation could significantly enhance the accuracy of lightweight models even in the absence of attention modules in the teacher model. The effectiveness of the developed models was then compared with model compression techniques such as quantization and pruning. Furthermore, this thesis provides the most comprehensive study to date on sensor-based HAR in resource-constrained environments, offering insights and strategies for the practical application of efficient and high-performance sensor-based HAR systems, thereby addressing a significant gap in the field.

BAP PhD Thesis Award for Dr. Alper Ahmetoğlu

Tue, 02 Jul 2024 00:00:00 +0000

Our doctoral graduate Alper Ahmetoğlu’s PhD dissertation titled “Neurosymbolic Representations for Lifelong Learning” has been awarded the DOCTORAL THESIS AWARD by the Boğaziçi University Scientific Research Projects Commission due to its outstanding scientific quality. Dr. Ahmetoğlu, a former member of the CoLoRs Lab, conducted his thesis under the supervision of Assoc. Prof. Dr. Emre Uğur and co-supervision of Prof. Dr. Erhan Öztop, and is currently continuing his postdoctoral research at Brown University.

Congratulations Dr. Berrenur Saylam!

Tue, 02 Jul 2024 00:00:00 +0000

Machine Learning for Well-being Assessment: Exploring Digital Biomarkers

Abstract

This thesis investigates the potential of wearable sensors and self-reported questionnaires by exploring different factors of human well-being using digital biomarkers. The primary goal is to identify and validate these biomarkers, comparing them against traditional psychological studies. Various machine learning models are trained and validated for analyzing different well-being factors, including sleep, mental well-being (stress, anxiety, depression, positive and negative affect), and academic achievement. Two extensive datasets are utilized in these explorations: NetHealth, which is collected from more than 700 college students over 4 years, and Tesserae collected over a year of 757 office workers. Advanced techniques, incorporating the analysis of time-lagged data to capture temporal patterns and multitask learning, are employed to unravel complex relationships among well-being parameters. The research unfolds systematically, progressing from single well-being factor exploration to time-based and multi-task methodologies. This thesis emphasizes the importance of incorporating temporal dimensions and multi-task learning strategies for a more comprehensive understanding of well-being and the factors influencing it. Our findings offer valuable insights into the identification of reliable biomarkers and the relationships between various well-being aspects within two different target groups of university students and office workers. It aims to provide a new perspective, moving beyond single-factor exploration to enhance the comprehensiveness and applicability of well-being studies.

Congratulations Dr. Beytullah Yiğit!

Mon, 03 Jun 2024 00:00:00 +0000

A Security Framework for Software-Defined Networks

Abstract

Software-Defined Networking (SDN) emerges as a transformative technology, offering the flexibility and scalability crucial for modern digital services. While SDN presents opportunities to address security shortcomings in traditional networks, it also introduces new vulnerabilities, particularly in data and control plane communications.

Congratulations Dr. Nuriye Özlem Özcan Şimşek!

Tue, 21 May 2024 00:00:00 +0000

Genomic Data Analysis Using Machine Learning Methods For Disease and Disease-Gene Prediction

Abstract

Genomic diseases arise due to certain mutations or combinations of mutations in the DNA. This combination can be different for each patient and the effect of each mutation is different for the disease. In this study, we are focussing on the genomic causes of diseases. We defined two research problems. One is to detect the disease from the genetic code, represented as list of mutations. The second is to detect disease-gene associations.

SIU 2024 Alper Atalay Awards

Sun, 19 May 2024 00:00:00 +0000

Two of our graduate students, Merve Gül Kantarcı and Timoteos Onur Özçelik, won the second and third prizes for the best student paper at the Signal Processing and Communication Applications conference held in Tarsus between May 14-18.

The Alper Atalay Best Student Paper Award at the Signal Processing and Communications Applications Conference is given in memory of Alper Atalay, who graduated with honors from Boğaziçi University, Department of Electrical and Electronics Engineering in 1993 and lost his life in a tragic car accident on August 16, 1998.

Congratulations Dr. Binnur Görer!

Fri, 08 Mar 2024 00:00:00 +0000

A Tutoring Framework with Robotic and Virtual Agents for Requirements Elicitation Interview Training

Abstract

Requirements elicitation interviews are a widely adopted technique where the interview success depends on the interviewer’s preparedness and communication skills. Scaling practice sessions for many students is challenging due to the need for stakeholder involvement. To address this, we introduce a robotic tutor for requirements elicitation interview training. It has components to support both the interview phase, wherein students act as interviewers while the system assumes the role of an interviewee, and the feedback phase, during which the system assesses students’ performance and offers contextual feedback. We empirically evaluated the system with a group of graduate students. Building on this study, we introduce a generic architecture to leverage multiple technologies for the tutoring agent. It also features a behavioral feedback evaluator to enhance students’ soft skills. We demonstrate the architecture’s applicability through two implementations: with a physical robotic agent and a virtual voice-only agent. Through empirical studies with another group of graduate students, we found that the participants appreciated both systems. They demonstrated higher learning gains when trained with the robotic agent but found the voice-only agent more engaging and easier to use. Our findings indicate that each system has distinct benefits and drawbacks, suggesting a customizable approach for educators’ preferences and resources. Additionally, we explore large language models to generate interview scenarios, aiming to augment the interview tutoring system with various scenarios. We assessed the quality of generated scripts with language quality metrics and an expert judgment study and confirmed their applicability.

Turna Turkish Language Model

Tue, 06 Feb 2024 00:00:00 +0000

Congratulations Dr. Alper Ahmetoğlu

Wed, 31 Jan 2024 00:00:00 +0000

Neurosymbolic Representations for Lifelong Learning

Abstract

This thesis presents a novel framework for robot learning that combines the advantages of deep neural architectures in processing high-dimensional vectors with classical AI search techniques to bridge the gap between continuous sensorimotor data of the robot and domains consisting of finite entities. The aim is to convert information about the environment collected through interactions into an appropriate symbolic form on which a search tree can be built to reach a desired state. The framework consists of an encoder-decoder type of network with binarized activations in the bottleneck layer. The state of the environment, represented as a set of object features, is given to the encoder as input. The output is a discrete vector, treated as the object’s symbol, given to the decoder together with the action vector. The decoder predicts the effect observed by the agent due to the executed action. Once the network is trained, we can transform the continuously represented environment definition into symbolic vectors using the encoder. This allows us to build rules defining the transitions in the environment defined over these symbols. These rules can be translated into planning domain definition language (PDDL), allowing domain-independent off-the-shelf planners to be used to search for a goal state. Our experiments on tabletop object manipulation setups show that the system can learn appropriate symbols of the environment that allow it to build object towers with desired heights and complex object structures that require modeling the relations between objects by reasoning through the rules defined over the symbols learned in an unsupervised manner. As the framework is built with differentiable blocks, it affords appending recent advances in deep learning with ease, allowing it to be extensible in multiple directions.

CMPE Graduation Ceremony 2023!

Fri, 07 Jul 2023 00:00:00 +0000

The CMPE Graduation Ceremony 2023 was held on July 7. After the diploma ceremony on the south campus in the morning, the new graduates, their family members, and members of CMPE came together in the garden of the CMPE building in the afternoon.

We congratulate the new graduates of CMPE and their family members. As members of CMPE, we are so proud of you and honored to celebrate your graduation with you. We wish you the best of luck with your new beginnings!

Yavuz Köroğlu defended his Phd thesis

Mon, 25 Jul 2022 00:00:00 +0000

Underestimating the value of software testing had catastrophic results in recent history. Automated Test Generation (ATG) is an approach that aims to minimize the manual effort required for testing. This thesis aims to improve the effectiveness and performance of ATG approaches via Machine Learning (ML) based guidance, and focuses on Android Graphical User Interface (GUI) testing using Reinforcement Learning (RL), specifically. We propose four solutions, Q-learning Based Exploration (QBE), Test Case Mutation (TCM), Fully Automated Reinforcement LEArning Driven (FARLEAD), and FARLEAD2 test generators. QBE uses RL to crawl a set of applications and learns an action generation policy while exploring. Then, it uses this learned policy to either detect more unique crashes or cover more activities in new applications. TCM takes the tests QBE generates and replaces the well-behaving actions in those tests with bad-behaving ones to detect even more crashes. FARLEAD uses RL to learn how to verify a functional behavior that is given as a high-level test scenario in the form of a monitorable formal specification. FARLEAD learns by trial-and-error like QBE but it learns app-specific patterns instead of QBE’s app-generic patterns. To the best of out knowledge, FARLEAD is the first engine fully automating the functional testing of GUI applications. Finally, FARLEAD2 improves FARLEAD with Generalized Experience Replay (GER) and human-readable Staged Test Scenario (STS) language.Experimental results show that, QBE outperforms state-of-the-art test generators in crash detection and coverage. Furthermore, executing QBE first and then switching to TCM detects even more unique crashes. FARLEAD and FARLEAD2 expand the scope of automated testing to verifying functional behavior. Overall, these test generators elevate automated GUI testing closer to replacing manual GUI testing.

CMPE Graduation Ceremony 2022!

Wed, 06 Jul 2022 00:00:00 +0000

The CMPE Graduation Ceremony 2022 was held on July 6. After the diploma ceremony on the south campus in the morning, the new graduates, their family members and members of CMPE came together in the garden of the CMPE building in the afternoon.

We congratulate the new graduates of CMPE and their family members. As members of CMPE, we are so proud of you and honoured to celebrate your graduation with you. We wish you the best of luck with your new beginnings!

Arzucan Özgür received the Excellence in Research Award

Tue, 25 Jan 2022 00:00:00 +0000

Lale Akarun received the Excellence in Research Award

Mon, 24 Jan 2022 00:00:00 +0000

Stories on Computer Engineering

Congratulations Dr. Goshgar Can Ismayilov!

Decentralized Privacy-Preserving Collective and Multi-Objective Trading Protocols on Blockchain with Zero-Knowledge Proofs

Congratulations Dr. Barış Yamansavaşçılar

Efficient Orchestration Methods in Air Computing Using Deep Reinforcement Learning

Abstract

A EuroHPC Success Story

Congratulations Dr. Yiğit Yıldırım

Learning robotic navigation and manipulation primitives from demonstration

Abstract

Fatih Alagöz received the Excellence in Research Award

Congratulations Dr. Sümeyye Ağaç!

Enhancing Lightweight Models for Efficient Sensor-based Human Activity Recognition

Abstract

BAP PhD Thesis Award for Dr. Alper Ahmetoğlu

Congratulations Dr. Berrenur Saylam!

Machine Learning for Well-being Assessment: Exploring Digital Biomarkers

Abstract

Congratulations Dr. Beytullah Yiğit!

A Security Framework for Software-Defined Networks

Abstract

Congratulations Dr. Nuriye Özlem Özcan Şimşek!

Genomic Data Analysis Using Machine Learning Methods For Disease and Disease-Gene Prediction

Abstract

SIU 2024 Alper Atalay Awards

Congratulations Dr. Binnur Görer!

A Tutoring Framework with Robotic and Virtual Agents for Requirements Elicitation Interview Training

Abstract

Turna Turkish Language Model

Congratulations Dr. Alper Ahmetoğlu

Neurosymbolic Representations for Lifelong Learning

Abstract

CMPE Graduation Ceremony 2023!

Yavuz Köroğlu defended his Phd thesis

CMPE Graduation Ceremony 2022!

Arzucan Özgür received the Excellence in Research Award

Lale Akarun received the Excellence in Research Award