Recommender systems recommend items (e.g., movies, products, books) to users. In this thesis, we proposed two comprehensive and cluster-induced recommendation-based methods: Orthogonal Inductive Matrix Completion (OMIC) and Burst-induced Multi-armed Bandit (BMAB). Given the presence of side information, the first method is categorized as context-aware. OMIC is the first matrix completion method to approach the problem of incorporating biases, side information terms and a pure low-rank term into a single flexible framework with a well-principled optimization procedure. The second method, BMAB, is context-free. That is, it does not require any side data about users or items. Unlike previous context-free multi-armed bandit approaches, our method considers the temporal dynamics of human communication on the web and treats the problem in a continuous time setting. We built our models' assumptions under solid theoretical foundations. For OMIC, we provided theoretical guarantees in the form of generalization bounds by considering the distribution-free case: no assumptions about the sampling distribution are made. Additionally, we conducted a theoretical analysis of community side information when the sampling distribution is known and an adjusted nuclear norm regularization is applied. We showed that our method requires just a few entries to accurately recover the ratings matrix if the structure of the ground truth closely matches the cluster side information. For BMAB, we provided regret guarantees under mild conditions that demonstrate how the system's stability affects the expected reward. Furthermore, we conducted extensive experiments to validate our proposed methodologies. In a controlled environment, we implemented synthetic data generation techniques capable of replicating the domains for which OMIC and BMAB were designed. As a result, we were able to analyze our algorithms' performance across a broad spectrum of ground truth regimes. Finally, we replicated a real-world scenario by utilizing well-established recommender datasets. After comparing our approaches to several baselines, we observe that they achieved state-of-the-art results in terms of accuracy. Apart from being highly accurate, these methods improve interpretability by describing and quantifying features of the datasets they characterize.

In recent decades, there has been increasing interest in analyzing the behavior of complex systems. A popular approach for analyzing such systems is a network analytic approach where the system is represented by a graph structure (Wassermann&Faust 1994, Boccaletti et al. 2006, Brandes&Erlebach 2005, Vespignani 2018): Nodes represent the system’s entities, edges their interactions. A large toolbox of network analytic methods, such as measures for structural properties (Newman 2010), centrality measures (Koschützki et al. 2005), or methods for identifying communities (Fortunato 2010), is readily available to be applied on any network structure. However, it is often overlooked that a network representation of a system and the (technically applicable) methods contain assumptions that need to be met; otherwise, the results are not interpretable or even misleading. The most important assumption of a network representation is the presence of indirect effects: If A has an impact on B, and B has an impact on C, then A has an impact on C (Zweig 2016, Brandes et al. 2013). The presence of indirect effects can be explained by ”something” flowing through the network by moving from node to node. Such network flows (or network processes) may be the propagation of information in social networks, the spread of infections, or entities using the network as infrastructure, such as in transportation networks. Also several network measures, particularly most centrality measures, assume the presence of such a network process, but additionally assume specific properties of the network processes (Borgatti 2005). Then, a centrality value indicates a node’s importance with respect to a process with these properties. While this has been known for several years, only recently have datasets containing real-world network flows become accessible. In this context, the goal of this dissertation is to provide a better understanding of the actual behavior of real-world network processes, with a particular focus on centrality measures: If real-world network processes turn out to show different properties than those assumed by classic centrality measures, these measures might considerably under- or overestimate the importance of nodes for the actual network flow. To the best of our knowledge, there are only very few works addressing this topic. The contributions of this thesis are therefore as follows: (i) We investigate in which aspects real-world network flows meet the assumptions contained about them in centrality measures. (ii) Since we find that the real-world flows show considerably different properties than assumed, we test to which extent the found properties can be explained by models, i.e., models based on shortest paths or random walks. (iii) We study whether the deviations from the assumed behavior have an impact on the results of centrality measures. To this end, we introduce flow-based variants of centrality measures which are either based on the assumed behavior or on the actual behavior of the real-world network flow. This enables systematic evaluation of the impact of each assumption on the resulting rankings of centrality measures. While–on a large scale–we observe a surprisingly large robustness of the measures against deviations in their assumptions, there are nodes whose importance is rated very differently when the real-world network flow is taken into account. (iv) As a technical contribution, we provide a method for an efficient handling of large sets of flow trajectories by summarizing them into groups of similar trajectories. (v) We furthermore present the results of an interdisciplinary research project in which the trajectories of humans in a network were analyzed in detail. In general, we are convinced that a process-driven perspective on network analysis in which the network process is considered in addition to the network representation, can help to better understand the behavior of complex systems.

Data is the new gold and serves as a key to answer the five W’s (Who, What, Where, When, Why) and How’s of any business. Companies are now mining data more than ever and one of the most important aspects while analyzing this data is to detect anomalous patterns to identify critical patterns and points. To tackle the vital aspects of timeseries analysis, this thesis presents a novel hybrid framework that stands on three pillars: Anomaly Detection, Uncertainty Estimation, and Interpretability and Explainability. The first pillar is comprised of contributions in the area of time-series anomaly detection. Deep Anomaly Detection for Time-series (DeepAnT), a novel deep learning-based anomaly detection method, lies at the foundation of the proposed hybrid framework and addresses the inadequacy of traditional anomaly detection methods. To the best of the author’s knowledge, Convolutional Neural Network (CNN) was used for the first time in Deep Anomaly Detection for Time-series (DeepAnT) to robustly detect multiple types of anomalies in the tricky and continuously changing time-series data. To further improve the anomaly detection performance, a fusion-based method, Fusion of Statistical and Deep Learning for Anomaly Detection (FuseAD) is proposed. This method aims to combine the strengths of existing wellfounded statistical methods and powerful data-driven methods. In the second pillar of this framework, a hybrid approach that combines the high accuracy of the deterministic models with the posterior distribution approximation of Bayesian neural networks is proposed. In the third pillar of the proposed framework, mechanisms to enable both HOW and WHY parts are presented.

Deep learning has achieved significant improvements in a variety of tasks in computer vision applications with an open image dataset which has a large amount of data. However, the acquisition of a large number of the dataset is a challenge in real-world applications, especially if they are new eras for deep learning. Furthermore, the distribution of class in the dataset is often imbalanced. The data imbalance problem is frequently bottlenecks of the neural network performance in classification. Recently, the potential of generative adversarial networks (GAN) as a data augmentation method on minority data has been studied. This dissertation investigates using GAN and transfer learning to improve the performance of the classification under imbalanced data conditions. We first propose a classification enhancement generative adversarial networks (CEGAN) to enhance the quality of generated synthetic minority data and more importantly, to improve the prediction accuracy in data imbalanced condition. Our experiments show that approximating the real data distribution using CEGAN improves the classification performance significantly in data imbalanced conditions compared with various standard data augmentation methods. To further improve the performance of the classification, we propose a novel supervised discriminative feature generation method (DFG) for minority class dataset. DFG is based on the modified structure of Generative Adversarial Network consisting of four independent networks: generator, discriminator, feature extractor, and classifier. To augment the selected discriminative features of minority class data by adopting attention mechanism, the generator for class-imbalanced target task is trained while feature extractor and classifier are regularized with the pre-trained ones from large source data. The experimental results show that the generator of DFG enhances the augmentation of label-preserved and diverse features, and classification results are significantly improved on the target task. In this thesis, these proposals are deployed to bearing fault detection and diagnosis of induction motor and shipping label recognition and validation for logistics. The experimental results for bearing fault detection and diagnosis conclude that the proposed GAN-based framework has good performance on the imbalanced fault diagnosis of rotating machinery. The experimental results for shipping label recognition and validation also show that the proposed method achieves better performance than many classical and state-of-the-art algorithms.

In the increasingly competitive public-cloud marketplace, improving the efficiency of data centers is a major concern. One way to improve efficiency is to consolidate as many VMs onto as few physical cores as possible, provided that performance expectations are not violated. However, as a prerequisite for increased VM densities, the hypervisor’s VM scheduler must allocate processor time efficiently and in a timely fashion. As we show in this thesis, contemporary VM schedulers leave substantial room for improvements in both regards when facing challenging high-VM-density workloads that frequently trigger the VM scheduler. As root causes, we identify (i) high runtime overheads and (ii) unpredictable scheduling heuristics. To better support high VM densities, we propose Tableau, a VM scheduler that guarantees a minimum processor share and a maximum bound on scheduling delay for every VM in the system. Tableau combines a low-overhead, core-local, table-driven dispatcher with a fast on-demand table-generation procedure (triggered on VM creation/teardown) that employs scheduling techniques typically used in hard real-time systems. Further, we show that, owing to its focus on efficiency and scalability, Tableau provides comparable or better throughput than existing Xen schedulers in dedicated-core scenarios as are commonly employed in public clouds today. Tableau also extends this design by providing the ability to use idle cycles in the system to perform low-priority background work, without affecting the performance of primary VMs, a common requirement in public clouds. Finally, VM churn and workload variations in multi-tenant public clouds result in changing interference patterns at runtime, resulting in performance variation. In particular, variation in last-level cache (LLC) interference has been shown to have a significant impact on virtualized application performance in cloud environments. Tableau employs a novel technique for dealing with dynamically changing interference, which involves periodically regenerating tables with the same guarantees on utilization and scheduling latency for all VMs in the system, but having different LLC interference characteristics. We present two strategies to mitigate LLC interference: a randomized approach, and one that uses performance counters to detect VMs running cache-intensive workloads and selectively mitigate interference.