Social media is a significant factor in information distribution today, and this information landscape contains a lot of different posts that compete for the user’s attention. Different factors can help catch the interest of the user, and one of them is the headline of the message. The headline can be more or less eye-catching, which can make the reader more or less interested in interacting with the post. The theme of this study is the automatized creation of eye-catching headlines that stay truthful to the content of the articles using Automatic Text Summarization. The exact method used consisted of fine-tuning the BART model, which is an existing model for Text Summarization. Other papers have been written using different models to solve this problem with more or less success, however, none have used this method. It was deemed an interesting method as it is less time- and energy-consuming than creating and training a new model entirely from scratch and therefore could be easily replicated if the results were positive. The BartForConditionalGeneration model implemented by the HuggingFace library was fine-tuned, using the Popular News Articles by Web.io. This method showed positive results. The resulting headlines were deemed faithful to the original ones, with a ROUGE-2 recall score of 0.541. They were comparably eye-catching to the human-written headlines, with the human respondents ranking them almost the same, with an average rank of 1.692 for the human-written headlines, and 1.821 for fine-tuned BART, and also getting an average score of 3.31 on a 1 to 5 attractiveness score scale. They were also deemed very comprehensible, with an average score of 0.95 on a scale from 0 to 1. / Sociala medier är idag en viktig faktor i distributionen av information. Detta nya landskap innehåller många olika inlägg som tävlar om användarens uppmärksamhet. Olika faktorer kan hjälpa till att fånga användarens blick till specifika inlägg eller artiklar, och en av dessa faktorer är rubriken. Rubriken kan vara mer eller mindre fängslande, och göra läsaren mer eller mindre intresserad av att interagera med inlägget. Temat för denna studie är att automatiskt skapa iögonfallande och intressanta rubriker, som beskriver innehå llet i artiklarna på ett korrekt sätt. Den valda metoden är automatisk textsamman fattning, och mer specifikt finjusterades BART-modellen, som är en existerande modell för textsammanfattning. Andra metoder har använts tidigare för att lösa denna problematik med mer eller mindre framgång, men ingen studie hade använt den här. Den ansågs vara intressant eftersom den är mindre tids- och energikrävande än vad det skulle vara att skapa en ny modell från grunden, och därför skulle den lätt kunna replikeras om resultatet var positivt. BartForConditionalGeneration-modellen implementerad av HuggingFace-bib lioteket finjusterades därför med hjälp av artiklar och rubriker från datasetet ’Popular News Articles’ av Web.io. Metoden visade positiva resultat. De resulterande rubrikerna ansågs trogna de ursprungliga, med en ROUGE-2 recall score på 0,541. De var jämförbart iögonfallande gentemot de mänskligt skrivna rubrikerna, då respondenterna rankade dem nästan likadant, med en genomsnittlig rankning på 1,692 för de mänskligt skrivna rubrikerna och 1,821 för rubrikerna som finjusterade BART genererade. De fick också ett genomsnittligt betyg av 3,31 på en poängskala från 1 till 5. De ansågs dessutom vara mycket lättbegripliga, med ett medelpoäng på 0,95 på en skala från 0 till 1.

In today's industrial power converters, converter reliability is essential, and converter topologies are well-established. Without a doubt, the power electronic industry continues to seek efficient power delivery and high power density. Resonant converters, especially LLC converters, have been intensively studied and applied in DC-DC converters. One of the most demanding applications for LLC converters is data centers. To date, LLC Resonant converters, are deployed in many applications for improved efficiency, density, and reliability. With the introduction of WBG devices coupled with the soft switching feature, the switching frequency can be extended beyond Mega-Hertz. With the significant increase in operating frequency, complicated magnetic components can be broken down into a cellular structure, each with a few number of turns. They can be easily implemented using 4-6 layers of PCB windings. Moreover, integrating the cellular cores using flux cancellation can further improve the power density. The proposed integrated magnetics can be automated in the manufacturing process. The magnetic size is reduced at this frequency, and planar magnetics using PCB winding become more relevant. PCB magnetics feature multiple advantages over Litz wire. The benefits are summarized as follows: The labor-intensive manufacturing process can be automated, thus reduction of cost. There is much reduced CM noise by using the shield layer. They have parasitics with much-improved reproducibility in large quantities. PCB windings feature less leakage between transformer windings because of the flexibility of the winding interleaving and the reduced number of turns. There is better thermal management due to the increased surface-to-body ratio. The design has a low profile and high-power density. However, it is not without its own limitations. There are challenges for high frequency PCB-magnetic magnetic design for the LLC converter. Firstly, With the recently developed high frequency core material, a phenomenon referred to as the dimensional resonant is observed. The effects of dimensional resonance were discussed in the literature when using an unusually large core structure; however, it can be observed more frequently under high excitation frequency, particularly with integrated magnetics. This dissertation discusses the dimensional effects of core loss on a PCB-based magnetics structure. A case study is presented on a 3-kW 400-to-48-V LLC prototype running at 1 MHz. The converter utilizes a low-profile matrix of two integrated transformers with a rectangular and thin cross-section area for reduced core loss. Specific solutions are presented. % Secondly, The matrix transformer is suitable for an LLC converter with high output current. However, the matrix transformer also increases the core size and core losses. The core loss degrades the LLC converter's light load and peak efficiency. In this dissertation, We discuss the design process and implementation of the DC-DC stage of the power supply unit for narrow range 48 V data center bus architecture. The optimization takes into account the number of elemental transformers, number of transformer turns, switching frequency, and transformer dimensions, namely winding width and core cross-section area. The optimization process results in a nearly 99% efficient 400-to-48-V LLC with a very high-power density and low profile fully integrated on PCB. A matrix of four transformers is used to reduce the termination loss of the secondary synchronous rectifier and achieve better thermal management. The number of secondary turns is optimized to achieve the best trade-off between winding loss, core loss, and power density. Another challenge arises for magnetic integration when multiple magnetic components with different characteristics come together. For instance, in the case of a transformer and an inductor on the same PCB. The PCB transformer is designed with perfectly interleaved primary and secondary layers to utilize the full PCB layer thickness. As a rule of thumb, the transformer winding layer is designed within 1 to 2 times the skin depth. On the other hand, the inductor's winding lacks interleaving and suffers from high MMF stress on layers. This makes the inductor prone to high eddy currents and eddy loss. Furthermore, this dissertation addresses the challenges associated with the high winding and core loss in the Integrated Transformer-Inductor (ITL). To overcome these challenges, we propose an improved winding design of the ITL by utilizing idle shielding layers for inductor integration within the matrix transformer. This method offers full printed circuit board (PCB) utilization, where all layers are consumed as winding, resulting in a significant reduction in the winding loss of the ITL. Moreover, we propose an improved core structure of the ITL that offers better flux distribution of the leakage flux within the magnetic core. This method reduces the core loss by more than 50% compared to the conventional core structure. We demonstrate the effectiveness of our proposed concepts by presenting the design of the ITL used in a high-efficiency, high-power-density 3-kW 400-to-48-V LLC module. The proposed converter achieves a peak efficiency of 98.7% and a power density of 1500 W/in3. This dissertation presents the concept of matrix inductors to solve such problems. A matrix of four resonant inductors is also designed to reduce the proximity effect between inductor windings and reduce inductor PCB winding loss. The matrix inductor provides a solution for high thermal stress in PCB-based inductors and reduces the inter-winding capacitance between inductor layers. This dissertation solves the challenges in magnetic design in high-frequency DC-DC converters in offline power supplies and data centers. This includes the transformer and inductor of the LLC converter. With the academic contribution in this dissertation, Wide-bandgap devices WBG can be successfully utilized in high-frequency DC-DC converters with Mega-Hertz switching frequency to achieve high efficiency, high power density, and automated manufacturing. The cost will be reduced, and the performance will be improved significantly. / Doctor of Philosophy / Industrial power converters need to be reliable and efficient to meet the power industry's demand for efficient power delivery and high power density. Research should focus on improving existing converter designs to improve fabrication, efficiency, and reliability. Resonant converters have been found to be effective in power conversion, especially in data centers where energy consumption is high. Three-element Resonant converters (LLC) are already used to improve efficiency, density, and reliability. By using Wide Bandgap devices and soft switching, the switching frequency can be extended beyond MHz, simplifying magnetic components and improving power density. The proposed integrated magnetics can be automated during the manufacturing process, further improving power density. At higher frequencies, planar magnetic components made with PCB winding are more effective than Litz wire. They are cheaper to make because of automation, have less common-mode noise, and are more reproducible in large quantities. PCB winding also has a low profile, high-power density, and better thermal management. However, it is not without its own limitations. There are challenges for high frequency PCB-magnetic magnetic design for the LLC converter. Firstly, With the recently developed high frequency core material, a phenomenon referred to as the dimensional resonant is observed. The effects of dimensional resonance were discussed in the literature when using an unusually large core structure; however, it can be observed more frequently under high excitation frequency, particularly with integrated magnetics. This dissertation discusses the effects of core loss on a PCB-based magnetics structure and presents solutions, including a case study on a 3-kW 400-to-48 V LLC prototype running at 1 MHz. Another challenge arises for magnetic integration when multiple magnetic components with different characteristics come together. For instance, in the case of a transformer and an inductor on the same PCB. The PCB transformer is designed with perfectly interleaved winding and low Ohmic loss. On the other hand, the inductor's winding lacks interleaving and suffers from a high proximity field. This makes the inductor prone to high eddy currents and eddy loss. This dissertation presents the concept of matrix inductors to solve such problems. A matrix of four resonant inductors is also designed to reduce the proximity effect between inductor windings and reduce inductor PCB winding loss. The matrix inductor provides a solution for high thermal stress in PCB-based inductors and reduces the inter-winding capacitance between inductor layers. Furthermore, this dissertation addresses the challenges associated with the high winding and core loss in the Integrated Transformer-Inductor (ITL). To overcome these challenges, we propose an improved winding design of the ITL by utilizing idle shielding layers for inductor integration within the matrix transformer. This method offers full printed circuit board (PCB) utilization, where all layers are consumed as winding, resulting in a significant reduction in the winding loss of the ITL. Moreover, we propose an improved core structure of the ITL that reduces the core loss by more than 50% compared to the conventional core structure. We demonstrate the effectiveness of our proposed concepts on a high-efficiency, high-power-density 3-kW 400-to-48-V LLC module. The proposed converter achieves a peak efficiency of 98.7% and a power density of 1500 W/in3. This dissertation solves the challenges in magnetic design in high-frequency DC-DC converters in offline power supplies and data centers. This includes the transformer and inductor of the LLC converter. With the academic contribution in this dissertation, Wide-bandgap devices WBG can be successfully utilized in high-frequency DC-DC converters with Mega-Hertz switching frequency to achieve high efficiency, high power density, and automated manufacturing. The cost will be reduced, and the performance will be improved significantly.

Chess is a deep and intricate game, the master of which depends on learning tens of thousands of the patterns that may occur on the board. At Noctie, their mission is to aid this learning process through humanlike chess AI. A prominent challenge lies in curating instructive chess positions for students. Usually these are either manually found by going through large numbers of real games, or handcrafted – a time-consuming process. For effective learning, it is often useful to collect many positions following the same theme, or exhibiting the same type of pattern. Curating such collections from real games is a challenging task. This thesis investigates the transformer decoder-only architecture and its capability of generating realistic, game-like chess-positions. This investigation involved the development and training of a decoder model using Pytorch, and a simple web-based Turing test gaining larger understanding of testers experience. The developed chess model successfully generates chess positions, with constraining possibilities of fixed pieces, score intervals, and fixed empty positions. Controlled re-generation ensures satisfaction of score intervals, while empty positions are handled by iterating over the model's probabilities. Based on the limited data provided by the Turing test, the model seems to fool players below 2000 rank-points on chess.com, where guess percentages land near the 50 percent mark, providing no clear indication that it deviates from randomness.

Over the past few years, the surge in VR (Virtual Reality) video traffic on networks has been remarkable. Nonetheless, a key challenge remains: ensuring a top-notch quality of experience (QoE) for VR video playback, especially when network bandwidth is limited. Prior studies have mainly focused on tile-based adaptive bitrate (ABR) streaming operating at the application layer on the server/client side to improve QoE, using single viewport prediction to conserve bandwidth. However, single-viewpoint prediction models face limitations due to uncertainties linked with head movement, making it difficult to handle sudden user motions effectively. To overcome these constraints, we propose a lightweight multimodal spatial-temporal transformer architecture, which generates multiple viewpoint trajectories and their corresponding probabilities while leveraging historical trajectory information. Consequently, we introduce a multi-agent reinforcement learning (MARL)-based ABR algorithm that capitalizes on multiple viewport prediction for VR video streaming at the application layer. Our algorithm strives to optimize various QoE objectives under diverse network conditions. To address the ABR problem, we formulate it as a Decentralized Partially Observable Markov Decision Process (Dec-POMDP) problem. To tackle this effectively, we develop a MAPPO (Multi-Agent Proximal Policy Optimization) algorithm within a centralized training and decentralized execution (CTDE) framework. Meanwhile, we also improve QoE at the network layer by utilizing network resources in different network nodes during VR video streaming. We present an innovative system called tile-weighted rate-distortion (TWRD) packet scheduling optimization, which takes advantage of viewpoint prediction. The system dynamically assigns weights to tiles and their corresponding packets using the probability of viewpoint prediction. Due to limited bandwidth, the problem of packet scheduling arises, requiring the determination of which packets should be dropped. To address this challenge, we formulate the problem as an optimization task, taking into account error propagation in the video. Our system leverages the weighted rate-distortion information of packets and applies dynamic programming techniques to design an optimal packet scheduling scheme. By selectively dropping packets at network nodes, our proposed system effectively reduces network congestion and enhances the overall performance of VR video streaming systems operating within bandwidth limitations.

The move toward electric vehicles (EVs) has a significant impact to reduce greenhouse gas (GHG) emissions and make transportation more eco-friendly. Fast-charging stations play a crucial role in this transition, making EVs more convenient for adoption specifically when driving in long distance. However, the challenge is to create a fast-charging system that can work with the different types of EVs and their varying power needs while still being efficient and effective. In this context, this thesis embarks on this journey by introducing an innovative solution for efficient universal fast charging, spanning both low voltage and high voltage battery systems. A novel, configurable dual secondary resonant converter is proposed, which empowers the charging module to extend its output range without imposing additional demands on the resonant tank components. This solution addresses the pressing need for a wide output voltage range in fast-charging standard in the growing EV landscape. To ensure optimal performance across a broad voltage and power range, the thesis employs an analytical model for LLC resonant converters to optimize the resonant components. This strategic component selection aims to achieve the desired output voltage and power range while minimizing conduction losses. The proposed topology and design methodology are rigorously validated through the development of a 10 kW prototype. Furthermore, the study introduces a two degrees of freedom (2DoF) control scheme for the proposed LLC resonant converter with the configurable dual secondary LLC converter topology. An analytical model is formulated to guide the selection of control parameters, ensuring coverage of the desired output voltage and power range without compromising system efficiency. The steady-state analytical model is utilized for determining optimized control parameters at each operating point within the converter's output range. To enhance the charging module's power density and efficiency, a high-frequency litz-wire transformer design methodology is introduced. The transformer's core size is optimized to achieve high power density and efficiency, while the winding configuration is chosen to minimize conduction losses. Finite Element Analysis (FEA) simulations validate transformer losses and operating temperatures. The culmination of this research is the development of a 30 kW charging module prototype. This prototype features an LLC resonant converter with a configurable dual secondary and two degrees of freedom control for output voltage control. The component ratings, estimated losses, and power board design are carefully considered to create a compact and efficient charging module. Experimental testing across a universal output voltage and power range con rms the effectiveness of the proposed solution. In summary, this thesis presents a comprehensive approach to design of a module for EV fast charging application addressing voltage range, efficiency, and component optimization, resulting in the successful development of a high-performance charging module prototype. / Thesis / Doctor of Engineering (DEng)

Molecular property prediction has a vast range of applications in the chemical industry. A powerful molecular property prediction model can promote experiments and production processes. The idea behind this degree program lies in the use of transfer learning to predict molecular properties. The project is divided into two parts. The first part is to build and pre-train the model. The model, which is constructed with pure attention-based Transformer Layer, is pre-trained through a Masked Edge Recovery task with large-scale unlabeled data. Then, the performance of this pre- trained model is tested with different molecular property prediction tasks and finally verifies the effectiveness of transfer learning.The results show that after self-supervised pre-training, this model shows its excellent generalization capability. It is possible to be fine-tuned with a short period and performs well in downstream tasks. And the effectiveness of transfer learning is reflected in the experiment as well. The pre-trained model not only shortens the task- specific training time but also obtains better performance and avoids overfitting due to too little training data for molecular property prediction. / Prediktion av molekylers egenskaper har en stor mängd tillämpningar inom kemiindustrin. Kraftfulla metoder för att predicera molekylära egenskaper kan främja vetenskapliga experiment och produktionsprocesser. Ansatsen i detta arbete är att använda överförd inlärning (eng. transfer learning) för att predicera egenskaper hos molekyler. Projektet är indelat i två delar. Den första delen fokuserar på att utveckla och förträna en modell. Modellen består av Transformer-lager med attention- mekanismer och förtränas genom att återställa maskerade kanter i molekylgrafer från storskaliga mängder icke-annoterad data. Efteråt utvärderas prestandan hos den förtränade modellen i en mängd olika uppgifter baserade på prediktion av molekylegenskaper vilket bekräftar fördelen med överförd inlärning.Resultaten visar att modellen efter självövervakad förträning besitter utmärkt förmåga till att generalisera. Den kan finjusteras med liten tidskostnad och presterar väl i specialiserade uppgifter. Effektiviteten hos överförd inlärning visas också i experimenten. Den förtränade modellen förkortar inte bara tiden för uppgifts-specifik inlärning utan uppnår även bättre prestanda och undviker att övertränas på grund otillräckliga mängder data i uppgifter för prediktion av molekylegenskaper.

Increased power quality problems due to intensive usage of power electronic devices resulted in development of software applications to perform quick harmonic analysis. However, the present harmonic analysis applications have special software or computer locks requirements and occupy huge memory and cost high. An application program (using Microsoft Visual C++) that is simple yet accurate in calculations; with no special software or high memory requirements is developed in this thesis work. The program uses the automatic acceptance criteria (AAC) and the harmonic penetration techniques in calculating the system voltages. Several userriendly features and tools that aid in better understanding of system harmonics are included in the program. Comparison of case study results with Superharm simulation results proves the program?s accuracy. This thesis work resulted in an informative and time saving program with which the user can document the study results and analyze them with minimum effort.

No description available.

No description available.

<p>A novel planar waveguide Time Domain Optical Fourier Transformer (TD-OFT), which is composed of waveguide lenses and blazed phase gratings, is proposed. A detailed mathematical derivation based on scalar diffraction optics is presented. In order to verify the theoretical analysis, the reciprocity in TD-OFT is also studied. Three different pulse examples, including the Gaussian pulse, square pulse and square pulse train, are implemented by analytical formulations. To evaluate the device performance, the similarity coefficient is defined. The results show that the similarity increases as the device aperture increases. However, there is trade-off between the similarity and the spectra resolution. For the input pulse, under the circumstance of same similarity, the shorter temporal pulse duration (larger bandwidth) needs smaller aperture size. Improved waveguide lens is particularly designed and then the whole device is simulated by Extension of BPM (EX-BPM) with two specific pulses, Gaussian and raised cosine pulse. The simulation results are also verified by reciprocity theorem using the numerical method. The designed TD-OFT occupies a size about 600μm (in width)×5mm (in length) for an ultrafast pulse around 10fs. It is possible to make the device size even smaller either by reducing the focal length of the collimating lens or enlarging the bandwidth of input pulse. Compared with currently proposed TD-OFT made by discrete photonic and optoelectronic components, this design can be integrated with a more compact size and seems more appealing on the simulated performance and fabrication cost. As a result, the planar waveguide TD-OFT has great potential in the next ultrafast optical network.</p> / Master of Applied Science (MASc)