Computer Engineering / Bilgisayar Mühendisliği
Permanent URI for this collectionhttps://hdl.handle.net/11147/10
Browse
34 results
Search Results
Now showing 1 - 10 of 34
Article Citation - Scopus: 3Cut-In Maneuver Detection With Self-Supervised Contrastive Video Representation Learning(Springer, 2023) Nalçakan, Yağız; Baştanlar, YalınThe detection of the maneuvers of the surrounding vehicles is important for autonomous vehicles to act accordingly to avoid possible accidents. This study proposes a framework based on contrastive representation learning to detect potentially dangerous cut-in maneuvers that can happen in front of the ego vehicle. First, the encoder network is trained in a self-supervised fashion with contrastive loss where two augmented videos of the same video clip stay close to each other in the embedding space, while augmentations from different videos stay far apart. Since no maneuver labeling is required in this step, a relatively large dataset can be used. After this self-supervised training, the encoder is fine-tuned with our cut-in/lane-pass labeled datasets. Instead of using original video frames, we simplified the scene by highlighting surrounding vehicles and ego-lane. We have investigated the use of several classification heads, augmentation types, and scene simplification alternatives. The most successful model outperforms the best fully supervised model by ∼ 2% with an accuracy of 92.52%Article Label-Free Retraining for Improved Ground Plane Segmentation(Springer, 2022) Uzyıldırım, Furkan Eren; Özuysal, MustafaDue to increased potential applications of unmanned aerial vehicles over urban areas, algorithms for the safe landing of these devices have become more critical. One way to ensure a safe landing is to locate the ground plane regions of images captured by the device camera that are free of obstacles by deep semantic segmentation networks. In this paper, we study the performance of semantic segmentation networks trained for this purpose at a particular altitude and location. We show that a variation in altitude and location significantly decreases network performance. We then propose an approach to retrain the network using only a new set of images and without marking the ground regions in this novel training set. Our experiments show that we can convert a network’s operating range from low to high altitudes and vice versa by label-free retraining.Article Asking the Right Questions To Solve Algebraic Word Problems(TÜBİTAK - Türkiye Bilimsel ve Teknolojik Araştırma Kurumu, 2022) Çelik, Ege Yiğit; Orulluoğlu, Zeynel; Mertoğlu, Rıdvan; Tekir, SelmaWord algebra problems are among challenging AI tasks as they combine natural language understanding with a formal equation system. Traditional approaches to the problem work with equation templates and frame the task as a template selection and number assignment to the selected template. The recent deep learning-based solutions exploit contextual language models like BERT and encode the natural language text to decode the corresponding equation system. The proposed approach is similar to the template-based methods as it works with a template and fills in the number slots. Nevertheless, it has contextual understanding because it adopts a question generation and answering pipeline to create tuples of numbers, to finally perform the number assignment task by custom sets of rules. The inspiring idea is that by asking the right questions and answering them using a state-of-the-art language model-based system, one can learn the correct values for the number slots in an equation system. The empirical results show that the proposed approach outperforms the other methods significantly on the word algebra benchmark dataset alg514 and performs the second best on the AI2 corpus for arithmetic word problems. It also has superior performance on the challenging SVAMP dataset. Though it is a rule-based system, simple rule sets and relatively slight differences between rules for different templates indicate that it is highly probable to develop a system that can learn the patterns for the collection of all possible templates, and produce the correct equations for an example instance.Article Citation - WoS: 7Citation - Scopus: 12A Survey on Organizational Choices for Microservice-Based Software Architectures(TÜBİTAK, 2022) Ünlü, Hüseyin; Bilgin, Burak; Demirörs, OnurDuring the last decade, the demand for more flexible, responsive, and reliable software applications increased exponentially. The availability of internet infrastructure and new software technologies to respond to this demand led to a new generation of applications. As a result, cloud-based, distributed, independently deployable web applications working together in a microservice-based software architecture style have gained popularity. The style has been a common practice in the industry and successfully utilized by companies. Adopting this style demands software organizations to transform their culture. However, there is a lack of research studies that explores common practices for microservices. Thus, we performed a survey to explore the organizational choices on software analysis, design, size measurement, and effort estimation when working with microservices. The results provide a snapshot of the software industry that utilizes microservices. We provide insight for software organizations to transform their culture and suggest challenges researchers can focus on in the area.Article Citation - WoS: 8Citation - Scopus: 9Dementia diagnosis by ensemble deep neural networks using FDG-PET scans(Springer, 2022) Yiğit, Altuğ; Baştanlar, Yalın; Işık, ZerrinDementia is a type of brain disease that affects the mental abilities. Various studies utilize PET features or some two-dimensional brain perspectives to diagnose dementia. In this study, we have proposed an ensemble approach, which employs volumetric and axial perspective features for the diagnosis of Alzheimer’s disease and the patients with mild cognitive impairment. We have employed deep learning models and constructed two disparate networks. The first network evaluates volumetric features, and the second network assesses grid-based brain scan features. Decisions of these networks were combined by an adaptive majority voting algorithm to create an ensemble learner. In the evaluations, we compared ensemble networks with single ones as well as feature fusion networks to identify possible improvement; as a result, the ensemble method turned out to be promising for making a diagnostic decision. The proposed ensemble network achieved an average accuracy of 91.83% for the diagnosis of Alzheimer’s disease; to the best of our knowledge, it is the highest diagnosis performance in the literature.Article Citation - Scopus: 1Performance Analysis and Feature Selection for Network-Based Intrusion Detection With Deep Learning(Türkiye Klinikleri, 2022) Caner, Serhat; Erdoğmuş, Nesli; Erten, Yusuf MuratAn intrusion detection system is an automated monitoring tool that analyzes network traffic and detects malicious activities by looking out either for known patterns of attacks or for an anomaly. In this study, intrusion detection and classification performances of different deep learning based systems are examined. For this purpose, 24 deep neural networks with four different architectures are trained and evaluated on CICIDS2017 dataset. Furthermore, the best performing model is utilized to inspect raw network traffic features and rank them with respect to their contributions to success rates. By selecting features with respect to their ranks, sets of varying size from 3 to 77 are assessed in terms of classification accuracy and time efficiency. The results show that recurrent neural networks with a certain level of complexity can achieve comparable success rates with state-of-the-art systems using a small feature set of size 9; while the average time required to classify a test sample is halved compared to the complete set.Article Citation - WoS: 1Citation - Scopus: 2Information Retrieval-Based Bug Localization Approach With Adaptive Attribute Weighting(TÜBİTAK - Türkiye Bilimsel ve Teknolojik Araştırma Kurumu, 2021) ErşahIn, Mustafa; Utku, Semih; Kılınç, Deniz; ErşahIn, BuketSoftware quality assurance is one of the crucial factors for the success of software projects. Bug fixing has an essential role in software quality assurance, and bug localization (BL) is the first step of this process. BL is difficult and time-consuming since the developers should understand the flow, coding structure, and the logic of the program. Information retrieval-based bug localization (IRBL) uses the information of bug reports and source code to locate the section of code in which the bug occurs. It is difficult to apply other tools because of the diversity of software development languages, design patterns, and development standards. The aim of this study is to build an adaptive IRBL tool and make it usable by more companies. BugSTAiR solves the aforementioned problem by means of the adaptive attribute weighting (AAW) algorithm and is evaluated on four open-source projects which are well-known benchmark datasets on BL. One of them is BLIA which is the state of the art in bug localization area and another is BLUIR which is a well-known BL tool. According to the promising results of experiments, Top1 rank of BugSTAiR is 2% and MAP is 10% better than BLIA's results on AspectJ and it has localized 4.6% of all bugs in Top1 and its precision is 6.1% better than BLIA on SWT, respectively. On the other side, it is 20% better in the Top1 metric and 30% in precision than BLUIR.Article Citation - WoS: 15Citation - Scopus: 18Achieving Query Performance in the Cloud Via a Cost-Effective Data Replication Strategy(Springer, 2021) Tos, Uras; Mokadem, Riad; Hameurlain, Abdelkader; Ayav, TolgaMeeting performance expectations of tenants without sacrificing economic benefit is a tough challenge for cloud providers. We propose a data replication strategy to simultaneously satisfy both the performance and provider profit. Response time of database queries is estimated with the consideration of parallel execution. If the estimated response time is not acceptable, bottlenecks are identified in the query plan. Data replication is realized to resolve the bottlenecks. Data placement is heuristically performed in a way to satisfy query response times at a minimal cost for the provider. We demonstrate the validity of our strategy in a performance evaluation study.Article Citation - WoS: 9Citation - Scopus: 11Test Input Generation From Cause-Effect Graphs(Springer, 2021) Kavzak Ufuktepe, Deniz; Ayav, Tolga; Belli, FevziCause-effect graphing is a well-known requirement-based and systematic testing method with a heuristic approach. Since it was introduced by Myers in 1979, there have not been any sufficiently comprehensive studies to generate test inputs from these graphs. However, there exist several methods for test input generation from Boolean expressions. Cause-effect graphs can be more convenient for a wide variety of users compared to Boolean expressions. Moreover, they can be used to enforce common constraints and rules on the system variables of different expressions of the system. This study proposes a new mutant-based test input generation method, Spectral Testing for Boolean specification models based on spectral analysis of Boolean expressions using mutations of the original expression. Unlike Myers' method, Spectral Testing is an algorithmic and deterministic method, in which we model the possible faults systematically. Furthermore, the conversion of cause-effect graphs between Boolean expressions is explored so that the existing test input generation methods for Boolean expressions can be exploited for cause-effect graphing. A software is developed as an open-source extendable tool for generating test inputs from cause-effect graphs by using different methods and performing mutation analysis for quantitative evaluation on these methods for further analysis and comparison. Selected methods, MI, MAX-A, MUTP, MNFP, CUTPNFP, MUMCUT, Unique MC/DC, and Masking MC/DC are implemented together with Myers' technique and the proposed Spectral Testing in the developed tool. For mutation testing, 9 common fault types of Boolean expressions are modeled, implemented, and generated in the tool. An XML-based standard on top of GraphML representing a cause-effect graph is proposed and is used as the input type to the approach. An empirical study is performed by a case study on 5 different systems with various requirements, including the benchmark set from the TCAS-II system. Our results show that the proposed XML-based cause-effect graph model can be used to represent system requirements. The developed tool can be used for test input generation from proposed cause-effect graph models and can perform mutation analysis to distinguish between the methods with respect to the effectiveness of test inputs and their mutant kill scores. The proposed Spectral Testing method outperforms the state-of-the-art methods in the context of critical systems, regarding both the effectiveness and mutant kill scores of the generated test inputs, and increasing the chances of revealing faults in the system and reducing the cost of testing. Moreover, the proposed method can be used as a separate or complementary method to other well-performing test input generation methods for covering specific fault types.Article Citation - WoS: 4Citation - Scopus: 4Predicting the Soft Error Vulnerability of Parallel Applications Using Machine Learning(Springer, 2021) Öz, Işıl; Arslan, SanemWith the widespread use of the multicore systems having smaller transistor sizes, soft errors become an important issue for parallel program execution. Fault injection is a prevalent method to quantify the soft error rates of the applications. However, it is very time consuming to perform detailed fault injection experiments. Therefore, prediction-based techniques have been proposed to evaluate the soft error vulnerability in a faster way. In this work, we present a soft error vulnerability prediction approach for parallel applications using machine learning algorithms. We define a set of features including thread communication, data sharing, parallel programming, and performance characteristics; and train our models based on three ML algorithms. This study uses the parallel programming features, as well as the combination of all features for the first time in vulnerability prediction of parallel programs. We propose two models for the soft error vulnerability prediction: (1) A regression model with rigorous feature selection analysis that estimates correct execution rates, (2) A novel classification model that predicts the vulnerability level of the target programs. We get maximum prediction accuracy rate of 73.2% for the regression-based model, and achieve 89% F-score for our classification model.