LEE- Bilgisayar Mühendisliği-Doktora
Bu koleksiyon için kalıcı URI
Gözat
Son Başvurular
1 - 5 / 20
-
ÖgeHierarchical deep bidirectional self-attention model for recommendation(Graduate School, 2023-05-02)This study proposes a bidirectional recommendation model to tackle the user cold start problem. We can predict the middle item when a user has only a few user-item interactions and enrich their interaction set accordingly. By recursively repeating this process, we can obtain enough interactions to make accurate item recommendations to the user. For instance, a user may buy a few items from an e-commerce site but also purchase other items from elsewhere, leading to incomplete information about their preferences. The proposed bidirectional recommendation model can fill the user's interaction history gaps, enabling accurate item recommendations even with limited data. In this thesis, we aimed to develop a recommendation system that imitates the behavior of today's e-commerce users' online purchasing experience. Our approach emphasized practicality, as we aimed to create a system that could be implemented easily in real-world e-commerce platforms. In doing so, we focused on developing an approach that could handle a large number of users and items found on such platforms while still maintaining high performance. By prioritizing these factors, we aimed to create a recommendation system that could be effectively applied in real-world scenarios. For future work, exploring combinations of the suggested algorithms for both layers would be worthwhile. Furthermore, examining the impact of algorithms proposed for the first layer or the user's shopping history enrichment algorithm on different recommendation systems would be beneficial. Ultimately, the most significant improvement is the application of proposed hierarchical recommendation network to cross-domain recommendation problems.
-
ÖgeDirectional regularization based variational models for image recovery(Graduate School, 2022-08-19)This thesis explores how local directional cues can be utilized in image recovery. Our intent is to provide image regularization paradigms that encourage the underlying directionalities. To this end, in the first phase of the thesis work, we design direction-aware analysis-based regularization terms. We boost the structure tensor total variation (STV) functionals used in inverse imaging problems so that they encode directional priors. More specifically, we suggest redefining structure tensors to describe the distribution of the ``directional" first-order derivatives within a local neighborhood. With this decision, we bring direction-awareness to the STV penalty terms, which were originally imposing local structural regularity. We enrich the nonlocal counterpart of the STV in the same way, which were additionally imposing nonlocal image self-similarity beforehand. These two types of regularizers are used to model denoising and deblurring problems within a variational framework. Since they result in convex energy functionals, we also develop convex optimization algorithms by devising the proximal maps of our direction-aware penalty terms. With these contributions in place, the major barrier in making these regularizers applicable lies in the difficulty of estimating directional parameters (i.e., the directions/orientations, the dose of anisotropy). Although, it is possible to come across uni-directional images, the real-world images usually exhibit no directional dominance. It is easy to precisely estimate the underlying directions of uni-directional (or partially directional) images. However, arbitrary and unstable directions call for spatially varying directional parameters. In this regard, we propose two different parameter estimation procedures, each of which employs the eigendecompositions of the semi-local/nonlocal structure tensors. We also make use of total variation (TV) regularization in one of the proposed procedures and a filterbank of anisotropic Gaussian kernels (AGKs) in the other. As our image regularization frameworks require the guidance of the directional parameter maps, we use the term ``direction-guided" in naming our regularizers. Through the quantitative and the visual experiments, we demonstrate how beneficial the involvement of the directional information is by validating the superiority of our regularizers over the state-of-the-art analysis-based regularization schemes, including STV and nonlocal STV. In the second phase of the thesis, we shift our focus from model-driven to data-driven image restoration, more specifically we deal with transfer learning. As the target field, we choose fluorescence microscopy imaging, where noise is a very usual phenomenon but data-driven denoising is less applicable due to lack of the ground-truth images. In order to tackle this challenge, we suggest tailoring a dataset by handpicking images from unrelated source datasets. This selective procedure explores some low-level view-based features (i.e., color, isotropy/anisotropy, and directionality) of the candidate images, and their similarities to those of the fluorescence microscopy images. Based upon our experience on the model-driven restoration techniques, we speculate that these low-level characteristics (especially directions) play an important role on image restoration. In order to encourage a deep learning model to exploit these characteristics, one could embed them into the training data. In fact, we establish the possibility of offering a good balance between content-awareness and universality of the model by transferring only low-level knowledge and letting the unrelated images bring additional knowledge. In addition to training a feed-forward denoising convolutional neural network (DnCNN) on our tailored dataset, we also suggest integrating a small amount of fluorescence data through the use of fine-tuning for better-recovered micrographs. We conduct extensive experiments considering both Gaussian and mixed Poisson-Gaussian denoising problems. On the one hand, the experiments show that our approach is able to curate a dataset, which is significantly superior to the arbitrarily chosen unrelated source datasets, and competitive against the real fluorescence images. On the other hand, the involvement of fine-tuning further boosts the performance by stimulating the content-awareness, at the expense of a limited amount of target-specific data that we assume is available.
-
ÖgeArtificial intelligence based and digital twin enabled aeronautical AD-HOC network management(Graduate School, 2022-12-20)The number of passengers using aircraft has been increasing gradually over the following years. With the increase in the number of passengers, significant changes in their needs have been made. In-flight connectivity (IFC) has become a crucial necessity for passengers with the evolving aeronautical technology. The passengers want to connect to the Internet without interruption regardless of their location and time. The aeronautical networks attract the attention of both industry and academia due to these reasons. Currently, satellite connectivity and air-to-ground (A2G) networks dominate existing IFC solutions. However, the high installation/equipment cost and latency of the satellites reduce their efficiency. Also, the terrestrial deployment of A2G stations reduces the coverage area, especially for remote flights over the ocean. One of the novel solutions is the Aeronautical Ad-hoc Networks (AANETs) to satisfy the IFC's huge demand by also solving the defects of satellite and A2G connectivities. The AANETs are based on creating air-to-air (A2A) links between airplanes and transmitting packets over these connections to enable IFC. The AANETs dramatically increase the Internet access rates of airplanes by widening the coverage area thanks to these established A2A links. However, the mobility and atmospheric effects on AANETs increase the A2A link breakages by leading to frequent aircraft replacement and reducing link quality. Accordingly, the mobility and atmospheric effects create the specific characteristics for AANETs. More specifically, the ultra-dynamic link characteristics of high-density airplanes create an unstructured and unstable topology in three-dimensional space for AANETs. To handle these specific characteristics, we first form a more stable, organized, and structured AANET topology. Then, we should continuously enable the sustainability and mapping of this created AANET topology by considering broken A2A links. Finally, we can route the packets over this formed, sustained, and mapped AANET topology. However, the above-explained AANET-specific characteristics restrict the applicability of conventional topology and routing management algorithms to AANET by increasing its complexity. More clearly, the AANET specific characteristics make its management challenging by reducing the packet delivery success of AANET with higher transfer delay. At that point, artificial intelligence (AI)-based solutions have been adapted to AANET to cope with the high management complexity by providing intelligent frameworks and architectures. Although AI-based management approaches are widely used in terrestrial networks, there is a lack of a comprehensive study that supports AI-based solutions for AANETs. Here, the AI-based AANET can take topology formation, sustainability, and routing management decisions in an automated fashion by considering its specific characteristics thanks to learning operations. Therefore, AI-based methodologies have an essential role in handling the management complexity of this hard-to-follow AANET environment as they support intelligent management architectures by also overcoming the drawbacks of conventional methodologies. On the other hand, these methodologies can increase the computational complexity of AANETs. At that point, we propose the utilization of the Digital Twin (DT) technology to handle computational complexity issues of AI-based methodologies. Based on these, in this thesis, we aim to propose an AI-based and DT-enabled management for AANETs. This system mainly consists of four main models as AANET Topology Formation Management, AANET Topology Sustainability Management, AANET Topology Mapping Management, and AANET Routing Management. Here, our first aim is to form a stable, organized, and structured AANET topology. Then, we will enable the sustainability of this formed topology. We also continuously map the formed and sustained AANET topology to airplanes. Finally, the packets of airplanes are routed on this formed, sustained, and mapped AANET topology. We will create these four models with different AI-based methodologies and combine all of them under the DT technology in the final step. In the Topology Formation Management, we will propose a three-phased topology formation model for AANETs based on unsupervised learning. The main reason for proposing an unsupervised learning-based algorithm is that we have independently located airplanes with unstructured characteristics in AANETs before forming the topology. They could be considered as the unlabeled training data for unsupervised learning. This management model utilizes the spatio-temporal locations of aircraft to create a more stable, organized, and structured AANET topology in the form of clusters. More clearly, the first phase corresponds to the aircraft clustering formation, and here, we aim to increase the AANET stability by creating spatially correlated clusters. The second phase consists of the A2A link determination for reducing the packet transfer delay. Finally, the cluster head selection increases the packet delivery ratio in AANET. In the Topology Sustainability Management, we will propose a learning vector quantization (LVQ) based topology sustainability model for AANETs based on supervised learning. The main reason for proposing a supervised learning-based algorithm is that we already have an AANET topology before the A2A link breakage, and we can use it in supervised learning for training. Accordingly, we can consider the clusters in AANET topology as a pattern; then, we can find the best matching cluster of an aircraft observing A2A link breakages through pattern classification instead of creating topology continuously. This management model works in three phases: winning cluster selection, intra-cluster link determination, and attribute update to increase the packet delivery ratio with reduced end-to-end latency. In the Topology Mapping Management, we will propose a gated recurrent unit (GRU) based topology mapping model for AANETs. In topology formation, we create AANET topology in the form of clusters by collecting airplanes having similar features under the same set. In topology sustainability, we sustain the formed clustered-AANET topology with supervised learning. However, these formed and sustained AANET topologies must be continuously mapped to the clustered airplanes to notify them about the current situation. This procedure could be considered a part of sustainability management. Here, we continuously notify the airplanes with GRU at each timestamp about topological changes. This management model works in two main parts ad forget and update gates. In Routing Management, we propose a q-learning (QLR) based routing management model for AANETs. For this aim, we map the AANET environment to reinforcement learning. Here, the QLR-based management model aims to let the airplanes find their routing path through exploration and exploitation. Accordingly, the routing algorithm can adapt to the dynamic conditions of AANETs. In this management model, we adapt the Bellman Equation to the AANET environment by proposing different methodologies for its related QLR components. Accordingly, this model mainly consists of two main parts current state & maximum state-action determination and dynamic reward determination. Therefore, we execute the topology formation, sustainability, and routing management modules through unsupervised, supervised, and reinforcement learning-based algorithms. Additionally, we take advantage of neural networks in topology mapping management. After managing the topology and routing of AANETs with AI-based models, in the DT-enabled AANET management, we will support them with the DT technology. The DT can virtually replicate the physical AANET components through closed-loop feedback in real-time to solve the computational challenges of AI-based methodologies. Therefore, we will introduce the utilization of DT technology for the AANET orchestration and propose a DT-enabled AANET (DT-AANET) management model. This model consists of the Physical AANET Twin and Controller, including the Digital AANET Twin with Operational Module. Here, the Digital AANET Twin virtually represents the physical environment. Also, the operational module executes the implemented AI-based models. Therefore, in this thesis, we aim to propose an AI-based and DT-enabled management for AANETs. In this management system, we will first aim to propose AI-based methodologies for AANET topology formation, topology sustainability, topology mapping, and routing issues. Then, we will support these AI-based methodologies with DT technology. This proposed complete management model increased the packet delivery success of AANETs with reduced end-to-end latency.
-
ÖgeNovel data partitioning and scheduling schemes for dynamic federated vehicular cloud(Graduate School, 2022-08-23)In the last decade, many Intelligent Transport Systems (ITS) applications, that rely on the Internet cloud infrastructure have been deployed to provide services to clients in vehicular environments. However, the Internet cloud suffers from high latency in service access and intermittent Internet disconnection in vehicular environments. In response to these challenges, researchers have proposed the use of other distributed computing technologies such as mobile cloud computing, edge computing, and vehicular cloud computing which may rely on Wireless Local Area Networks (WLAN) or Ad-hoc networks for communication to provide computing resources and services in vehicular environments. Vehicular cloud computing, an emerging distributed computing paradigm that provides cloud services using the resources embedded in vehicles and Road Side Units (RSUs), is arguably the best alternative to other distributed computing systems because of its advantages. The provision of vehicular cloud services leads to efficient utilization of the abundant resources in embedded vehicles and therefore embraces green technology in vehicular environments. Considering that modern vehicles and RSUs have been designed with communication capabilities through the use of the Vehicular Ad-hoc Network (VANET), vehicular cloud computing may not require additional communication hardware infrastructure installation along roads and at parking lots before its deployment. Furthermore, access to vehicular cloud services may involve a relatively lower latency and fewer intermittent disconnections due to the close proximity of vehicular cloud resources to clients and the use of VANET which may provide better connections among vehicular resources and clients than the connection to Internet cloud in vehicular environments. As with all emerging technologies of which vehicular cloud is not an exception, some challenges need to be addressed before its full real-world deployment. The challenges of vehicular cloud computing are mainly caused by the unique characteristics of vehicular resources and the limited communication bandwidth of VANET. In vehicular environments, i.e., on roads and parking lots, vehicles are not permanently stationed but mobile, making the resources embedded in vehicles highly mobile. Also, the resources of vehicles that are organized to provide cloud services belong to different people (distributed ownership). The high mobility and distributed ownership of resources imply that vehicular resources may exit the provisioned vehicular cloud abruptly by either withdrawal of resources by the resource owner or intermittent network disconnection caused by the high mobility of vehicles. Therefore distributed ownership of resources and high mobility of vehicles cause vehicular resources to be volatile, making their availability and reliability unpredictable. The capacity constraint of the communication bandwidth of VANET, which serves as a communication backbone of vehicular cloud, implies that the transmission of data-intensive and bandwidth-intensive applications in vehicular environments is a challenge in vehicular cloud computing. Considering that the distributed ownership of resources, limited communication bandwidth, and high mobility of vehicles lead to low availability and low reliability of resources in vehicular cloud computing, they adversely affect almost all resource management operations, such as virtual machine migration. Therefore, the main focus of this dissertation is to propose novel solutions to address the challenges of vehicular cloud and ameliorate the adverse effects of the identified characteristics of vehicular cloud computing. As an introduction, the background of vehicular cloud computing and a detailed survey of resource management operations in vehicular cloud are presented using a three-phase taxonomy of resource management techniques proposed in this dissertation. Based on the review of vehicular cloud computing concepts, a novel distributed vehicular computing paradigm, Vehicular Volunteer Computing (VVC), is proposed in this dissertation. VVC is a volunteer computing platform where vehicle owners may donate their idle processing units towards the execution of scientific and other projects that are beneficial to a community, such as the "Compute The Cure" cancer project. The concept of a Dynamic Federated Vehicular Cloud (DFVC) is introduced in this dissertation to overcome the challenge of limited resource capacity and volatility of resources in vehicular cloud. The DFVC entails the organization of resources from different vehicles moving on the road to provide a specific vehicular cloud service such as Computation-as-a-Service (CaaS). Although the resources are collated from different vehicles in the formation of DFVC, they are organized as a single logical unit for the provision of services. The formation of a DFVC involves forming resource-based clusters, i.e., grouping vehicles with similar resource and mobility characteristics as a single unit and selecting a leader, known as a cluster head, to manage the resources in a cluster of vehicles. By considering the structure of resource-based clusters formed in a Region of Interest (RoI), two different DFVC schemes are proposed in this dissertation: the Cluster-Based Vehicular Cloud (CBVC) and the Platoon-Based Vehicular Cloud (PBVC). In the CBVC, vehicles (owners) with a high reputation and idle resources on an RoI on the road are organized into clusters without adhering to the condition that all cluster members use the same lane and maintain a fixed gap between vehicles. The PBVC, on the other hand, is a variant of the cluster-based vehicular cloud that requires strict adherence to a constant gap, i.e., the distance between all vehicles and the use of the same lane throughout the entire period of the provision of cloud services. In other words, the PBVC is a convoy with a constant gap between all vehicles whose resources are organized to offer specific cloud services. In order to address the challenge of limited computation capacity of resources and constraint communication bandwidth, a large divisible data load to be processed by the DFVC is partitioned and distributed to the individual vehicular nodes using efficient Data partitioning and Scheduling (DPS) schemes. One of the central themes of this dissertation is, therefore, to design and implement novel DPS schemes that consider the characteristics of vehicular resources of the DFVC and the communication channel of VANET: the communication backbone for DFVC. By considering the computation capacity of resources, data transmission bandwidth capacity, and communication delay experienced in data transmission in VANET, efficient DPS schemes proposed in this dissertation are designed through mathematical models developed using timing and data flow diagrams. The DPS schemes for the CBVC, and PBVC are modeled differently because of their unique characteristics and operations. The proposed DPS scheme for the CBVC was modeled with the consideration that the cluster head determines the data chunk for each vehicle using derived closed form mathematical equations and then distributes the data chunks directly as a single hop to the respective vehicles in the CVBC to process. After processing, the vehicles then transmit the processed data chunks directly to the cluster head. The DPS scheme for the CBVC is implemented as part of a unified data, resource, and channel management framework, which is referred to as the UniDRM in this dissertation. Considering different criteria or objectives for data partitioning and scheduling, three distinct DPS schemes, time-aware, cost-aware, and reliability-aware, are also proposed in this dissertation. For the PBVC, the data partitioning is carried out by the lead node, i.e., the first node of the platoon, using derived closed form equations. The determined data chunks of the platoon members are then distributed either directly (single hop) or through multi-hop transmission. The closed form equations were derived considering data flow and timing diagrams designed based on how vehicles in platoons exchange information with their neighbor nodes, which is referred to as platoon Information Flow Topologies (IFT). In this dissertation, existing platoon IFTs: the Bi-Directional (BD), Bi-Directional Lead (BDL), and the All- to- All IFT (A2A) are modified to derive mathematical models for six different DPS schemes, namely, the Bi-Directional-Recursive (BD-R), Bi-Directional Interlaced (BD-I), Bi-Directional Lead-Recursive (BDL-R), Bi-Directional Lead-Interlaced (BDL-I), Bi-Directional Lead Aggregate-Recursive (BDLA-R), and Bi-Directional Lead Aggregate-Interlaced (BDLA-I). Through realistic simulations developed via the use of the simulation platforms Omnet++, Sumo, Veins, and Plexe, a detailed performance analysis of the proposed DPS schemes were carried out. By developing the different DPS schemes for the PBVC, one of the long-standing challenges of divisible data partitioning and scheduling in the literature: the modeling and derivation of closed form equations for determining the percentage of data chunks and the processing time of the linearly arranged network of connected heterogeneous processors, is addressed. According to the literature, there are no closed form equations for the heterogeneous linear network of connected processors because of the complicated combinatorial terms that appear in expressions for individual data partitions while solving the recursive equations. However, using algebraic manipulations, closed form equations have been derived and modeled in this dissertation. In all, this dissertation presents novel solutions to key challenges of vehicular cloud computing, including the limited available capacity of resources in vehicles and the bandwidth constraint of the vehicular communications channel.
-
ÖgeStatistical feature learning and signal generation for time-series sensor signals(Graduate School, 2022-05-31)The Human Activity Recognition (HAR) problem has attracted substantial attention from academia. HAR has many applications like smart home assisted living systems, healthcare monitoring systems, sports activity monitoring, and monitoring indoor and outdoor activities. HAR applications involve advanced machine learning techniques to identify and classify human activities by leveraging video cameras, wearable sensors, or any other signal like Wi-Fi or radar which eventually encodes the human activity. Human activities are encoded in signals and signal processing techniques are required to pre-process raw signals to filter out high-frequency components and to frame the signals into the fixed-length window. Wearable smart electronics are widely used in human daily life. Those smart devices contain sensors like accelerometer and gyroscope to measure triaxial acceleration and angular velocity respectively. Smartwatches, smartphones, or any such wearable sensor devices contain out-of-the-box sensors embedded in the device. Identification and classification of human activities from such signals by leveraging machine learning techniques require features to be extracted from the signal which represents the corresponding human activity. Many feature extraction techniques from such time-series signals exist in the literature. Time and frequency domain-based feature extraction is a widely used technique for sensor-based human activity classification. To train deep learning models, one needs features to be extracted from the signal. Though time and frequency domain feature extraction techniques are very efficient, the selection of the time and frequency domain features may have a significant impact on the overall classification accuracy. Alternatively, energy-based generative models eliminate the need for a feature extraction layer in the learning pipeline. Deep Belief Networks are alternatives to deep learning models eliminating the need for time and frequency-based feature extraction for sensor-based human activity classification: Restricted Boltzmann Machines (RBM) are the building blocks of Deep Belief Networks. RBMs are energy-based probabilistic graphical models which factorize the probability distribution of a random variable over a binary probability distribution. The visible layer of RBMs represents the real-valued random variable and the hidden layer represents the corresponding binary valued probability distribution. Conditional Restricted Boltzmann Machine (CRBM) is an extension to RBMs and is strong in capturing temporal dependency information encoded in time-series signals. They can be used in the classification of sensor-based human activities. The capacity of CRBM by factorizing a real-valued random variable probability distribution over a binary valued probability distribution eliminates the need for feature extraction from the signal by applying certain feature extraction techniques. This work shows how CRBM is trained to learn signal features. Once trained the signal is generated and reconstructed by the trained model. Along with CRBM, the results of other generative models RBM, GAN, WGAN-GP, and predictive model LSTM are also presented. To compare the performance of the models, similarity metrics are used as a performance criterion to show the performance of the generative models in generating the signals closest to the real signals. Euclidean, Canberra, and Dynamic Time Warping (DTW) distances are used as performance criteria. The results indicate that CRBM outperforms GAN, WGAN-GP, and RBM generative models in generating the signal closest to the original signal. LSTM performs close to CRBM. The capacity of the CRBM in generating signals closest to the original signal indicates that CRBM can learn features from the signal and can also be used in supervised classification.