Payment channels allow parties to utilize the blockchain to send transactions for a cheaper fee. Previous work has analyzed to which degree a party can profit by facilitating the transaction process. The aim is to increase the usability of the network and to be rewarded for providing this service. However, previous work focuses on maximizing the reward of the individual player in isolation, a model that we aim to expand. That is why in this work we extend the action space to allow other parties to act and react, and observe the impact this has on the rewards of the player that would otherwise act in isolation.
Testing existing placement strategies by performing channel placement games, we can assess the difference in the reward that indicates the potential loss that competition may cause when operating in the Bitcoin Lightning Network.
Furthermore, we have developed a new strategy that is able to improve the performance in the multi-actor model.

Federated Learning is a machine learning paradigm for decentralized training over different clients. The training happens in rounds where each client learns a specific model which is then aggregated by a central server and passed back to the clients. Since the paradigm’s inception, many frameworks that provide Federated Learning tools and infrastructure have appeared. This leads to the question of ”How do different Federated Learning frameworks compare?”, which is the research question of this paper. The paper’s main contribution will be helping developers new to the Federated Learning field decide between NVidia Flare, OpenFL, and Flower, three popular federated learning frameworks.

Federated learning enables the construction of machine learning models, while adhering to privacy constraints and without sharing data between different devices. It is achieved by creating a machine learning model on each device that contains data, and then combining these models through an aggregation algorithm without sharing the data. Federated learning is currently a hot topic, and a lot of research has gone into implementing accurate aggregation algorithms. The original algorithm is FedAvg, and since then many different algorithms have been introduced. In this paper, I will compare the performance of five different aggregation algorithms: FedAvg, FedProx, FedYogi, FedMedian and q-FedAvg. The algorithms are compared on different data sets, namely MNIST and a kinase inhibition data set, as well as on different data distributions and number of clients. The experiments indicate that among these five algorithms, FedYogi achieves the best performance, both in terms of highest final accuracy as well as in terms of convergence rate.

Data processing systems have become increasingly important in modern computing, as the volume and complexity of data that needs to be analyzed has grown dramatically. Multiple data processing systems have been and are being developed, that are scalable, resilient and performant.

However, despite the advances made in data processing technology, there are still challenges that need to be addressed in order to optimize the performance, energy efficiency as well as the practical- ity of these systems. One such challenge is the need to effectively manage the underlying system’s resources, including the system’s throughput and the amount of work that each operator has to do and to use optimal data-structures that would lead in faster task processing speeds.

To address this challenge, this thesis proposes the implementation of a high-performance data processing system that exposes the underlying system’s metrics to the application level and applys an innovative way for operator communication, by utilizing an efficient thread-safe data structure. By providing underlying system’s metrics to the application’s scheduler, the scheduler can schedule the tasks optimally according to the current system’s state and adjust the system’s resources during run- time. This alleviates the developers from having to fine-tune the system beforehand and allows the system to tackle fluctuating input workload more efficiently.

This thesis will explore the design and implementation of such system, as well as its impact on the performance, energy-efficiency and resiliency of data processing applications. We provide perfor- mance measurements as well as a qualitative comparison of our system compared to other state-of-the art systems, proving our hypotheses.

Blockchain technology has proven to be a promising solution for decentralized systems in various industries. At the core of a blockchain system is the peer-to-peer (P2P) overlay, which facilitates communication be- tween parties in the blockchain system. Recently, there is increasing evidence that this P2P overlay plays a major role in limited transaction throughput, limited scalability, and security threats in a blockchain system. In this work, it is explained that this bottleneck is caused by the fact that currently deployed P2P overlays cannot fully capture the complexity of a blockchain system. We propose a novel approach to address these issues by introducing a novel overlay architecture. Our approach, Sovereign Domains P2P Overlays, allows for nodes in a single network to belong to multiple heterogeneous groups (called domains), each with their own set of protocols tailored to the characteristics and needs of the nodes in that domain. To demonstrate the effectiveness of the proposed overlay architecture, two novel node discovery protocols, FedKad and SovKad, were designed and implemented. The simulations conducted in a custom-built simulator show that SovKad outperforms node discovery in a Structured P2P Overlay (Kademlia) and node discovery in a Federated P2P Overlay (FedKad), providing evidence of the potential of the Sovereign Domains P2P Overlay architecture in blockchain systems.

Large­scale machine learning frameworks can accelerate training of a neural network by per­ forming distributed training on a cluster using multiple GPUs per node and multiple nodes. Because distributed training on a cluster involves many nodes which need to communicate and load and exchange data, a machine learning framework may at certain times during training not fully utilize the available hardware of the system. Various techniques are as­ sessed in their capability to measure the performance of specific parts of the hardware of a cluster. We present ML Board, a tool that measures and visualizes the utilization of the system while training a neural network model using some of the previously assessed tech­niques, and does so without requiring any changes to the used machine learning framework. ML Board can be used to identify straggling nodes, and by subsequently letting the user select different nodes using the Slurm job scheduler, can help to decrease the training time of a ResNet model by between 15 and 45% when using an ImageNet or CIFAR­10 dataset. Furthermore, the energy used by the GPUs can be measured and used to identify and replace GPUs to reduce the total used energy by between 5 to 16%.

Modern systems generate a tremendous amount of data, making manual investigations infeasible, hence requiring automating the process of analysis. However, running automated log analysis pipelines is far from straightforward, due to the changing nature of software ecosystems caused by the constant need to adapt to user requirements. In practice, these are comprised of a series of steps that collectively aim at turning raw logs into actionable insights. The first step is log parsing which aims to abstract away from raw logs toward structured information. Log parsing is paramount, as it influences the performance of all subsequent downstream tasks that rely on its output. Although previous works have investigated the performance of log parsing, given the increase in data heterogeneity witnessed over the past decades, the validity of current estimates is questionable, as there is a lack of understanding of how log parsing methods perform in modern contexts. Consequently, we investigate the field and, in the process, we discover that misleading metrics are adopted, which produce incomplete performance estimates. Furthermore, motivated by an industry use case within the infrastructure of a large international financial institution, we discover that the current log parsing paradigm is not aligned with what is required in practice. Consequently, to address these current limitations, in this work we contribute with the following. We (1) evaluate the field of log parsing, (2) propose a new log parsing paradigm and create a benchmark dataset to facilitate future research, and (3) propose and evaluate a machine learning model that solves log parsing within the new paradigm.

One of the key problems of swarm robotics is how the mobile robots can navigate accurately in a given environment. To achieve this, the mobile robots need to accurately determine where they are globally, or relative to other robots and landmarks. This paper is going to be an investigation of the Monte Carlo Localization algorithm in an environment containing 3 anchors. In addition, an alternative modification is suggested to this localization algorithm to improve the localization performance. An experiment is devised to assess different aspects of localization performances of these algorithms. With the experimental results, a quantitative analysis of performance figures are compared and analyzed.

In this paper, we explore the creation of control algorithms for swarms of robots playing the role of either predator or prey in an environment filled with static obstacles. The paper devel- ops on a famous flock simulation model proposed by Craig Reynolds called boids. The paper analyzes a zero-sum game situation, in which one swarm of robots, the prey, is trying to reach a certain pre-determined target, while another swarm of robots, the predator, is trying to prevent it from reaching its objective. Swarm control algorithms for both the predator and the prey scenarios are analyzed in an arms race manner. The robots are modelled as point-mass holonomic entities, that can move in arbitrary directions. The proposed algorithms are tested on characteristics such as success rate and time in a simulated environment. As a result, a set of algorithms for both the predator and prey are proposed and their strength and weaknesses are discussed.

A majority of existing single-anchor localization algorithms make use of antenna arrays or special antenna systems. However, the need for specialized antenna systems incurs higher costs, complexity and power consumption. This paper presents a novel single-anchor localization algorithm, which does not require antenna arrays or special antennas. The algorithm is implemented in a two-robot system, where one of the robots acts as the anchor and the other acts as the target. The localization algorithm uses velocity measurements of each robot and distance measurements between robots. Using the change in distance between the robots and the velocity of the target robot relative to the anchor robot, the target robot can be localized relative to the anchor robot. The localization algorithm uses a Kalman filter as a state estimator of the movement of the target robot and a Savitzky-Golay filter to filter the distance measurements. A simulation with Gaussian noise in the distance measurements is performed to display the impact of the filtering methods. Another simulation showed an average localization error of 1.38 meters when the distance measurements are acquired through RSSI measurements, which is comparable to existing RSSI-based localization methods.

Robotic swarms provide a great many uses within a world increasingly relying on autonomous systems. Alas these swarms are also very vulnerable to faults, even the smallest fault can cripple the performance of a whole swarm. Such a fault could be one of several types; those that have to do with sensors, communication, ... This paper will try to discover the effect that these faults have on the performance of a robotic swarm attempting to complete a specified task, e.g. a multiple destination routing problem or a variant of a foraging task. With the experiments concluded and the data gathered it proved possible to observe correlations between fault and performance, but this dependency heavily changes based on the algorithm used and fault introduced.

The workflow of a data science practitioner includes gathering information from different sources and applying machine learning (ML) models. Such dispersed information can be combined through a process known as Data Integration (DI), which defines relations between entities and attributes. When all information is combined in one source suited for ML purposes, this source often contains duplicate data, resulting in longer operation times. Recent work has created algebraic rewrite rules for ML such that computations can be pushed down to individual sources. This method is referred to as factorized learning. In this work, we present an implementation of Amalur, a system for automated factorized learning that is novel because of its applicability in any DI scenario. Amalur shows significant speedups for some datasets and models, while for some other datasets and models, our implementation of factorized learning results in slowdowns. Whether factorized learning is efficient depends on the operations involved in the applied ML model and the data supplied to the model. The process of estimating the efficiency of factorization is known as cost estimation. Previous efforts on cost estimation do not generalize to the DI scenarios included in Amalur, are not tailored to different ML models, and are not properly evaluated. In this thesis, we create a cost estimation procedure for Amalur suitable for any DI scenario, which adapts to individual ML models. To evaluate this procedure we create a data generator capable of generating customizable DI scenarios. Our method outperforms its competitor on DI scenarios covered by the state of the art, and shows comparable performance on newly covered DI scenarios.

Current speed of data growth has exponentially increased over the past decade, highlighting the need of modern organizations for data discovery systems. Several (automated) schema matching approaches have been proposed to find related data, exploiting different parts of schema information (e.g. data type, data distribution, column name, etc.). However, research showed that single schema matching techniques fails to effectively match schemas, whilst combinatorial schema matching systems show more promise. With the introduction of combinatorial schema matching systems, new challenges arise regarding selection and combining strategies. This research attempts to explore different techniques for determining the importance of each matcher in a combinatorial schema matching system by determining the weights of each matcher and comparing them through a comprehensive evaluation.

Search engines play an important role in the provision of information. Recently, researchers have raised their concerns about the potential of search engine providers trying to shift the opinion of their users by manipulating search results. But search results might also be manipulated for financial gain. If an increase in User Engagement (UE) can be achieved by manipulating search results, the manipulation of search results can lead to increased advertisement revenue. In this thesis, a 2 x 3 x 3 factorial user study (n = 376) investigating the effect of biased Search Engine Results Pages (SERPs) on UE in users with strong opinions on the topic is presented. Different search result ranking biases were used: a search result ranking biased to agree with the topic, one that disagrees with the topic and one that is indifferent towards the topic. One of two different topics is assigned to the user. The bias in the SERP is determined by the condition assigned to the user (i.e., mitigate, control or reinforce) and the viewpoint of the user towards the topic (i.e., agree or disagree). Furthermore, one out of three queries is presented at the top of the SERP (i.e., a query that agrees, disagrees or is impartial with the topic). It was found that neither mitigating nor reinforcing bias through the SERP affected UE. The exploratory analysis suggests that confirmation bias did not guide the results clicked by the users. Users were not refrained from clicking search results that do not align with their viewpoint on the topic. Instead, the exploratory analysis suggests that results are clicked based on their position (position bias), regardless of whether bias of the user is mitigated or reinforced.

Recent years have shown a tremendous increase in the application of Artificial Intelligence to the field of radiology, often through the extraction and analysis of large numbers of quantitative features from medical images. These applications increase the demand for machine learning models to extract information from these images. To provide these models, improve their performance and reduce the time that experts have to spend on manually tuning them, the field of Automated Machine Learning (AutoML) aims to automate the design process of machine learning models by optimizing the selection of algorithms and their hyperparameters for each application. This work applies an AutoML approach to medical image classification, using a Bayesian optimization strategy to automatically optimize the selection of preprocessing and classification algorithms and their hyperparameters. Its performance is compared with the performance of a random search optimization strategy, evaluated on three datasets from three different clinical applications. The results show that the Bayesian optimization and the random search return models that achieve similar performance on the unseen test sets. We show that a random search with relatively few evaluations and a simple ensemble strategy is sufficient to achieve performance comparable to a more sophisticated and more computationally demanding Bayesian optimization approach, therefore validating the use of a random search optimization strategy in this medical image classification setting. All found models generalize poorly, with average F1-scores on the validation sets used for optimizing the models being at least 20\% lower than the average F1-scores on the unseen test sets. Finally, we further emphasize the difficulty to generalize in this setting, by showing that the differences between subsets of the evaluated datasets are large and that increasing the computation time of the optimization does not benefit the test set performance of the final solution.

While multiprocessor platforms have been widely adopted by the embedded systems industry in the past couple of years, there are still fundamental challenges about their timing predictability for applications with real-time timing constraints. The common-off-the-shelf (COTS) multiprocessor platforms typically use complex hardware components, interconnects and multi-level caches which are designed to deliver higher average-case performance. However, these features negatively impact the worst-case performance as they increase the interference of tasks on shared hardware resources. One effective software-based solution to counteract these issues is to use the non-preemptive execution model.

Despite its positive impact on timing predictability, non-preemptive execution causes a potential blocking problem which can decrease the ability to guarantee all timing constraints of the system. It is also known that scheduling non-preemptive periodic tasks on multiprocessor platforms is an NP-hard problem.

In this thesis, we focus on non-preemptive execution of sequential as well as parallel real-time tasks upon multiprocessor platforms and investigate, extend, and improve the state of the art on global, partitioned, and semi-partitioned scheduling approaches for the problem.

We provide the first necessary test for partition-ability, i.e., a test that can determine whether a given task set cannot be partitioned on a given number of cores regardless of the partitioning policy. This test allows us to quantify the pessimism of the existing partitioning heuristics as well as obtain the limits of partitioned scheduling.
We further introduce the first non-work-conserving global scheduling policy and show that despite the fact that it improves over the existing global scheduling policies, it is not as effective as the partitioned scheduling strategies.

We extend a sustainable scheduling algorithm designed for uni-processor platforms to multiprocessor ones to improve the performance of partitioning heuristics. A sustainable scheduling algorithm does not have timing anomalies and hence it is easier to analyze and can have better scheduling results.

Furthermore, we introduce the first semi-partitioned non-preemptive scheduling solution for multiprocessor platforms. Our solution is able to schedule some of the task sets for which it is impossible to find a partitioning solution. Finally, we compare the overheads and memory consumption of various scheduling approaches (including ours) on a bare-metal multiprocessor hardware platform, i.e., a 4 processor Raspberry Pi board. We show that our sustainable scheduler has a very low overhead while it out-performs other solutions in terms of schedulability.

Large cities in the Netherlands, like Rotterdam, have hundreds of playgrounds, but local governments have little information on how children and adults use them. Usage data can help create playgrounds that better fit with the residents' needs by identifying what elements of a playground are the most popular, and which do not get used. The AMS Institute and municipality of Rotterdam have asked us to develop a system to collect usage information of playgrounds without recording personally identifiable information. The system requirements are to record the locations of individual users and to estimate if a user is a child or an adult. It must do this without requiring an external power source and without a broadband internet connection. After considering different types of sensors (including computer vision, radio, sound, and mechanical acoustic signals), We decided to use a mmWave radar sensor due to its ability to provide accurate localization, easy installation, and low energy usage without recording identifiable information. We use a commercially available mmWave radar sensor that we configure to localize people in a 30m by 20m area when placed at the perimeter. We use the radar point cloud output from the device for classification by calculating statistics that we use as features for our classifier. We evaluate classifiers based on SVM, Random Forrest, fully connected and recurrent neural nets. We also analyze different methods for combining radar point clouds captured over time from the same person. We collected 100.000 radar point clouds of adults, children, and bicyclists at real playgrounds, split into a training and validation data set. We show that our SVM classifier achieves an accuracy of 79% on single radar frames, and 93% on the combined results of 10 second long sequences of our validation dataset. The classifier requires <1KB of memory and little processing power. Meaning it can execute on an embedded platform powered by a solar panel.