263134 research outputs found
Sort by
Developer Dialogues: A study of game creators to understand the potential for industry self-regulation of monetization
No full textCurrent scientific discourse on game monetization reflects strong ethical perspectives on the role of money in game production. The present study seeks to uncover the foundational principles and practices that shape how games are monetized. By interviewing 17 game creators directly involved in monetization, we uncover six principles and practices that might mitigate potential harms to player wellbeing associated with monetization. This research aims to inform ongoing efforts advocating for industry self-regulation concerning loot boxes and similar game elements.sponsorship: Fonds Wetenschappelijk Onderzoek|S006821Nstatus: Published onlin
Ondernemende handelaars in het globale Atlantische systeem: Frederic Romberg en de Oostenrijks- Nederlandse handel met West- en Centraal-Afrika, 1775-95
No full textThe historiography on the Atlantic System, of which the trade with Africa was part, has long been characterized by a focus on the major colonial powers and their central power strategies of empire-building, especially through mercantilist barriers and monopolistic companies. Recently, this traditional perspective has been challenged in numerous ways. First, historical strands such as Atlantic or global history have pleaded to look beyond nations and empires, describing the boundaries of Atlantic polities as spaces of transnational connection rather than separation. Second, scholars have reassessed the agency of private, enterprising merchants and their informal, international networks as a counterweight to the aforementioned focus on the monopolistic company. These novel perspectives have had several important consequences. First, they compounded existing Atlantic histories of nations in a rich way by including new cross-imperial and informal connections. Second, and more crucially, they also had the consequence of drawing attention to the participation in the colonial project of European areas not yet considered. Using these shifts in perspective, I examine the ties of the Southern Netherlands with the Atlantic complex using the case-study of Frederic Romberg, an eighteenth-century merchant settled in Brussels who moved into the slave trade during the American War of Independence (1775-83). I demonstrate how Romberg partnered with foreign slave traders, acquired slaving licences from the Spanish court, and forcibly transported thousands of captive Africans toward the Americas. By conducting a micro-study of the organization, strategies, and consequences of Romberg's operations, I dispel the idea that the Southern Netherlands were disconnected from the European colonial project.status: Publishe
Flexibele Piëzoelektrische Microgemachineerde Ultrageluid Transducers (PMUT) voor Beeldvormingstoepassingen
No full textUltrasound technology plays a crucial role in medical imaging, primarily using rigid probes with thickness mode piezoelectric vibration. However, limitations in fabrication techniques, such as mechanical dicing, have spurred the development of micromachined alternatives like Piezoelectric Micromachined Ultrasonic Transducers (PMUTs) and Capacitive MUTs (CMUTs), which offer compactness, lower power consumption, batch fabrication, and CMOS integration. Flexible ultrasound transducers can further expand medical imaging applications by providing conformal skin contact, user-friendliness for continuous monitoring, reduced operator variability, and improved angular coverage with power-efficient mechanical focusing. This PhD project, targeting at the intersection of the aforementioned ideas, develops a flexible PMUT towards imaging, addressing an unexplored area.
The research methodology employs a silicon-on-insulator (SOI)-free process and a custom-made lead zirconate titanate (PZT) solution to fabricate low-cost piezoelectric thin films via the sol-gel process and later, PMUT devices. The films, 1-2 µm thick, are characterized using X-ray diffraction (XRD) and double-beam laser interferometry (DBLI), yielding promising piezoelectric coefficients. Both rigid and flexible PMUTs are developed by depositing polyimide membranes and etching cavities, with flexible devices requiring additional polyimide deposition and a release step. Initial fully addressed arrays validated the fabrication process, which was later adapted to create row-column (RC)-addressed arrays with improved aperture and fill factors. Sensitivity and resonance frequencies were measured in air and water for performance comparison. Arrays are integrated with PCBs, connected to Verasonics systems, and waterproofed using polyimide, epoxy, and parylene-C. Imaging capabilities were demonstrated with rigid RC arrays capturing ultrasound images of wire and tissue-mimicking phantoms.
Main outcome of this PhD include: 1) Developing PZT solution recipe and characterizing sol-gel PZT. Using the solution as an active piezo layer, fabrication of rigid SOI, rigid-flex, and fully flex PMUTs with fully addressed and RC addressed designs. 2) Fabrication of PZT-based flexible RC PMUT for the first time with waterproof packaging and connection to electronics and underwater acoustic characterization (TX and RX) under bending. 3) Imaging with Rigid-flex RC PMUT (offering higher underwater sensitivity than flexible counterparts) demonstrated for the first time with interesting outcomes for low-frequency imaging.status: Publishe
Het verbeteren van niet-lineaire microgolfmeettechnieken voor de karakterisatie van actieve componenten en schakelingen
No full textThe evolution of cellular networks has revolutionized global communication, enabling digital connectivity, multimedia services, and Internet of Things applications with ultra-fast, reliable connections. Each network generation, from analog 1G to high-speed and cell-free 6G, has driven significant advancements, requiring transmitters to handle complex modulations, support increasing bandwidths, and meet rising efficiency requirements.
This PhD thesis introduces advanced nonlinear measurement and modeling techniques to tackle challenges in wireless network development, with a focus on load-pull systems and measurement-based modeling tools essential for optimizing microwave devices and circuits. These enhancements are crucial to improving RF and microwave systems that support new telecommunication standards.
A central aspect is the estimation of measurement uncertainty in load-pull systems, which have evolved thanks to the introduction of large-signal network analyzers and are now crucial for creating accurate models of active microwave devices, especially under nonlinear conditions. This work addresses gaps in uncertainty analysis by developing methods to quantify and mitigate errors, leading to more reliable measurement data for designing complex microwave circuits.
Additionally, this PhD thesis tackles behavioral modeling of power amplifiers under dynamic load modulation. Traditional models often lack precision for wideband, nonlinear scenarios, resulting in longer design cycles and performance discrepancies. The proposed approach overcomes these limitations by accurately modeling transistors and power amplifiers under dynamic load modulation, providing more realistic predictions and expediting design processes for wideband systems.
To streamline measurement processes, the research enhances the extraction of multitone X-parameters models. We introduce a method that reduces the number of frequencies at which small-signal excitation tones must be injected at the device terminals. By employing advanced interpolation methods, this approach preserves accuracy while minimizing measurement time, facilitating more efficient wideband modeling of microwave devices.
The PhD thesis also contributes to identifying the safe operating area of microwave devices during potentially destructive measurements. The research proposes an automated solution to monitor harmful conditions and estimate the risk of irreversible damage based on a constantly updated device model. This approach allows experimental campaigns to run without risking device failure, reducing the need for repeated measurements and lowering costs.
The final area of investigation explores combining the instantaneous bandwidths of multiple arbitrary waveform generators to produce a phase-coherent wideband signal. It is achieved by synchronizing the generators and adjusting phase shifts between their local oscillators. However, signal quality degrades over time due to phase drift, which limits its usability for extended measurement campaigns.
Overall, this research contributes to more precise, efficient, and safe measurement and modeling techniques in RF and microwave engineering, providing practical advancements for designing and testing microwave devices. These innovations are critical for progressing telecommunication technologies and setting future standards in RF engineering.status: Publishe
Just transitions, trade unions and the need for a new (European) social contract
No full textstatus: Publishe
Leerobjectieven voor Relationele Structuur in Deep Learning
No full textRelational structure is abundant both in the world and in our perception and contemplation of it, everything stands in relation to a multitude of other things. It would be impossible to gain real understanding of such a world by considering incidents or entities in isolation, without recognizing and exploiting those relations that are meaningful. Yet, deep learning systems are too often trained by doing exactly this, by being fed isolated samples, one by one, or by learning objectives that ignore underlying structure.
This dissertation proposes and studies learning objectives that leverage relational structure for deep learning methods. We argue for the need of effective representations that allow for adequate structure to be encoded, we discuss example applications where modeling relational structure proved effective, and we shed light on the lack of compositional understanding of current systems while proposing a potential solution. First, we argue that AI systems need representations that allow for reliable communication on one hand, and for a more structured, compositional reasoning process on the other hand. This could be achieved by using elements of discrete representations and processing techniques, by which we could more effectively encode into neural networks the relational structure we find in the world.
Then, we study how to leverage various kinds of relational structure of increasing complexity in several application domains. Imposing a sequential structure using learning-to-rank objectives, as well as appropriately modeling sentence interactions, improves an autoregressive system for a challenging textual retrieval task. Leveraging spatiotemporal relations leads to effective self-supervised pre-training objectives that allow us to learn informative representations for satellite imagery.
Next, we induce more complex structures that encode hierarchical syntax relations in a formal grammar for musical harmony, and we design an objective that emphasizes music-theoretical distances. Finally, we show that current models fail to generalize to unexpected situations when generating 2D visual scenes from text. Syntax improves compositional generalization in this context, if enforced by a learning objective that we propose in order to align the syntax structure with visual output representations. We envision that the use of hierarchical, recursive structure, which can be induced in an unsupervised manner, can contribute to a more compositional understanding and reasoning in future deep learning models.status: Publishe
De impact van ondermijnend geloof op traumatisch geheugen
No full textThe current research project is to examine the impact of undermining belief on traumatic memory. Specifically, three studies will be conducted. In Study 1, we will use a trauma film paradigm to examine whether belief in traumatic memories can be successfully reduced leading to nonbelieved memories. Study 2 will examine the effect of undermining belief and undermining nonbelieved memories on the creation of intrusive memories; Study 3 will investigate the effect of undermining belief and undermining nonbelieved memories on the memory.status: Publishe
Leren van Autonoom Rijden met Menselijk Gedrag uit Demonstraties: Een Differentieerbare Optimale Controle-Aanpak
No full textThe successful adoption of autonomous vehicles depends not only on technical advancements but also on user acceptance, demanding a human-centered approach to the design of automated driving functionalities. To incorporate the human perspective into their development, these systems can learn from demonstrations of preferred driving behaviors. An end-to-end approach can support this by jointly optimizing perception, planning, and control for the demonstrated driving task. However, such an approach may lack safety assurances and interpretability. In contrast, a modular approach offers greater interpretability and verifiability by optimizing each task independently, but it struggles to effectively learn driving behaviors from observation-action demonstrations.
Among the various techniques for autonomous driving, model predictive control (MPC) stands out as a popular model-based control method that computes control actions by optimizing the future states of the system over a time horizon while incorporating constraints on states and controls. However, MPC demands meticulous and time-consuming tuning of its cost function and may struggle with the flexibility required to imitate a wide range of human behaviors.
This thesis aims to integrate MPC with imitation learning, and evaluate its effectiveness in imitating human driving behaviors. To achieve this, it leverages differentiable optimal control, a technique that enables the computation of gradients of the optimal control actions with respect to MPC parameters, such as cost function terms. This approach allows to tune the MPC to match the driving demonstrations, while preserving its advantageous properties.
The thesis first addresses the development of a differentiable formulation of parametric MPC tailored for autonomous driving applications. This formulation, relying on the implicit differentiation of the optimality conditions of the problem, allows for nonlinear dynamics and state constraints. While learning with an MPC policy shows more robustness than a neural network policy, it proves less effective at generating closed-loop behaviors indistinguishable from human driving. To address this limitation, MPC needs to be integrated with other learning components, which are most effective when trained jointly with the MPC.
Building on these considerations, the thesis introduces the integration of differentiable MPC into an end-to-end architecture. This framework effectively maps high-dimensional data to safe and feasible low-level control actions, all without disrupting the end-to-end pipeline. As a result, it enables safer and more interpretable imitation learning, along with an online parameter tuning mechanism for MPC cost functions.
The methodologies presented in the thesis are validated through extensive simulation experiments involving human participants. A driving simulator, featuring a moving-based motion platform, is used to collect human driving demonstrations and capture various driving styles to be imitated. By training on these datasets, the proposed method outperforms other approaches involving MPC and neural networks, achieving an average 20% reduction in deviations from the demonstrated driving behaviors. Additionally, the effectiveness of the imitation is evaluated through a human-in-the-loop experiment on the simulator, targeted at determining whether passengers can distinguish between human drivers and the autonomous driving systems learned from these demonstrations. The results show that controllers integrating MPC with end-to-end learning are statistically less distinguishable from human driving than standard MPC controllers.status: Publishe
Foto-elektrokatalytische afbraak van recalcitrante organische verontreinigende stoffen met behulp van niet-metalen katalysatoren
No full textVarious advanced oxidation processes (AOPs) have been developed to achieve efficient elimination of these pharmaceuticals from aquatic environments over the past few decades. Highly reactive species, such as hydroxyl radicals, sulfate radicals and superoxide anions, can be continuously generated to breakdown pharmaceutical pollutants. Among all AOPs, photocatalysis, electrocatalysis and photoelectrocatalysis have received increasing attention because of their highly efficient performance. Various types of photocatalysts, electrocatalysts and electrode materials have been synthesized and applied to remove pharmaceuticals. Most of these developed catalysts are made from noble metals, transition metals and their alloys and face the dilemmas of high cost and unsustainability because of their non-renewability. Many improvements have been made in the degradation of pharmaceutical organics from aquatic environments. Nevertheless, new concerns have arisen over the generation of toxic degradation intermediates originating from the incomplete mineralization of these compounds by AOPs. Ciprofloxacin (CIP) is one of the most frequently used antibiotics, and its presence has been widely detected in environmental matrices of surface water and wastewater. It has been reported that some degradation intermediates of CIP by AOPs, especially those with fluorine, retain the antibacterial activity of CIP and exhibit higher toxicity than CIP. In addition, fundamental questions related to CIP degradation, such as the roles of these reactive species and how they react with CIP molecules, remain unclear.
Consequently, this PhD research focused on developing a photoelectrocatalytic system with non-metal catalysts for the degradation and detoxification of pharmaceuticals (in particular CIP). An efficient non-metal photocatalyst, e.g., nitrogen-defective graphitic carbon nitride (g-C3N4), was synthesized via a simple defect engineering technique in Chapter 2. The photocatalytic activity of the obtained nitrogen-defective g-C3N4 was tuned by controlling the incorporated nitrogen vacancy density. An intricate defect state mechanism was proposed to decipher the profound impact of nitrogen vacancies according to solid experimental analysis and theoretical calculations. The obtained nitrogen-defective g-C3N4 exhibited significantly improved performance for CIP degradation under simulated solar light illumination, as described in Chapter 3. Density function theory (DFT) calculations were conducted to predict the preferential attack sites on CIP by reactive species to provide deeper insights into the CIP degradation and defluorination mechanisms. To avoid the external addition of chemical oxidants in classical AOPs, such as H2O2, an efficient carbonaceous electrode, e.g., nitric acid-treated reticulated vitreous carbon (HNO3-RVC), was fabricated in Chapter 4, and it performed well in the in situ generation of H2O2 and the removal of CIP simultaneously. Furthermore, a photoelectrocatalytic system was developed in Chapter 5 based on the obtained nitrogen-defective g-C3N4 from Chapter 2 and HNO3-RVC from Chapter 4 to promote its performance in the degradation and detoxification of CIP wastewater. Respiratory and rotifer experiments were conducted to monitor the toxicity evolution of the treated solution. Molecular dynamics simulations and DFT calculations were performed to elucidate how reactive species, such as hydroxyl radicals and singlet oxygen, react with CIP molecules. Finally, a valorization plan was developed to explore the opportunities for knowledge transfer of the research results obtained in this thesis, bridging the gap between lab experiments and real applications.status: Publishe
Existential and Humanistic Integrative Therapies
No full textOne of the current discussions in the field of psychotherapy is whether established psychotherapeutic paradigms still have a future. Existential–integrative therapy, with its specific applications such as anxiety disorders and depression in adolescents, or further integrated with other therapeutic modalities such as equine-facilitated therapy for people with anorexia nervosa, has widened the humanistic–existential approach in a culture and time where its practice might be under pressure. This chapter searches for criteria that help integrate new elements to not only broaden but also deepen humanistic and existential approaches. The chapter addresses clinical situations and societal changes of the past and the future that call for an integration of new ideas and practices. It presents existential–integrative therapy as a distinct example of humanistic and existential integrative practices. Finally, the chapter focuses on emergent perspectives and limitations of integration within and outside the humanistic and existential paradigm.status: Accepte