Deep Neural Networks are widely applied to various kinds of fields nowadays. However, hundreds of thousands of neurons in each layer result in intensive memory storage requirement and a massive number of operations, making it difficult to employ deep neural networks on mobile devices where the hardware resources are limited. One common technique to address the memory limitation is to prune and quantize the neural networks. Besides, due to the frequent usage of Rectified Linear Unit (ReLU) function or network pruning, majority of the data in the weight matrices will be zeros, which will not only take up a large amount of memory space but also cause unnecessary computation operations. In this thesis, a new value-based compression method is put forward to represent sparse matrix more efficiently by eliminating these zero elements, and a customized hardware is implemented to realize the decompression and computation operations. The value-based compression method is aimed to replace the nonzero data in each column of the weight matrix with a reference value (arithmetic mean) and the relative differences between each nonzero element and the reference value. Intuitively, the data stored in each column is likely to contain similar values. Therefore, the differences will have a narrow range, and fewer bits rather than the full form will be sufficient to represent all the differences. In this way, the weight matrix can be further compressed to save memory space. The proposed value-based compression method reduces the memory storage requirement for the fully-connected layers of AlexNet to 37%, 41%, 47% and 68% of the compressed model, e.g., the Compressed Sparse Column (CSC) format, when the data size is set to 8 bits and the sparsity is 20%, 40%, 60% and 80% respectively. In the meanwhile, 41%, 53% and 63% compression rates of the fully-connected layers of the compressed AlexNet model with respect to 8-bit, 16-bit and 32-bit data are achieved when the sparsity is 40%. Similar results are obtained for VGG16 experiment. / Djupa neurala nätverk används i stor utsträckning inom olika fält nuförtiden. Emellertid ställer hundratusentals neuroner per lager krav på intensiv minneslagring och ett stort antal operationer, vilket gör det svårt att använda djupa neurala nätverk på mobila enheter där hårdvaruresurserna är begränsade. En vanlig teknik för att hantera minnesbegränsningen är att beskära och kvantifiera de neurala nätverken. På grund av den frekventa användningen av Rectified Linear Unit (ReLU) -funktionen eller nätverksbeskärning kommer majoriteten av datat i viktmatriserna att vara nollor, vilket inte bara tar upp mycket minnesutrymme utan också orsakar onödiga beräkningsoperationer. I denna avhandling presenteras en ny värdebaserad komprimeringsmetod för att representera den glesa matrisen mer effektivt genom att eliminera dessa nollelement, och en anpassad hårdvara implementeras för att realisera dekompressions- och beräkningsoperationerna. Den värdebaserade komprimeringsmetoden syftar till att ersätta icke-nolldata i varje kolumn i viktmatrisen med ett referensvärde (aritmetiskt medelvärde) och de relativa skillnaderna mellan varje icke-nollelement och referensvärdet. Intuitivt kommer data som lagras i varje kolumn sannolikt att innehålla liknande värden. Därför kommer skillnaderna att ha ett smalt intervall, och färre bitar snarare än den fullständiga formen kommer att räcka för att representera alla skillnader. På så sätt kan viktmatrisen komprimeras ytterligare för att spara minnesutrymme. Den föreslagna värdebaserade komprimeringsmetoden minskar minneslagringskravet för de helt anslutna lagren av AlexNet till 37%, 41%, 47% och 68% av den komprimerade modellen, t.ex. Compressed Sparse Column (CSC) format, när datastorleken är inställd på 8 bitar och sparsiteten är 20%, 40%, 60% respektive 80%. Under tiden uppnås 41%, 53% och 63% komprimeringshastigheter för de helt anslutna lagren i den komprimerade AlexNet-modellen med avseende på 8- bitars, 16-bitars och 32-bitars data när sparsiteten är 40%. Liknande resultat erhålls för VGG16-experiment.

This thesis aims to find opportunities to support communication between new mothers to foster a sense of connectedness, during a pandemic where social interactions such as parent groups are unavailable. The project is based on research that shows that new parents have a great need for social support. Due to physical distancing, there was room for interaction design to intervene by exploring tangible solutions to develop innovative custom design. Through a user-centered design process, including findings from theory, interviews and design workshops with the target group, the study resulted in a co-created design proposal of a communication device in the form of a wearable wristband. The proposal showed positive indications in meeting the needs and desires of mothers found during a co-design workshop.

No description available.

No description available.

A l'heure où l'industrie des semi-conducteurs fait face à des difficultés majeures pour entretenir une croissance en berne, les nouveaux outils de synthèse de haut niveau repositionnent les FPGAs comme une technologie de premier plan pour l'accélération matérielle d'algorithmes face aux clusters à base de CPUs et GPUs. Mais en l’état, pour un ingénieur logiciel, ces outils ne garantissent pas, sans expertise du matériel sous-jacent, l’utilisation de ces technologies à leur plein potentiel. Cette particularité peut alors constituer un frein à leur démocratisation. C'est pourquoi nous proposons une méthodologie d'accélération d'algorithmes sur FPGA. Après avoir présenté un modèle d'architecture haut niveau de cette cible, nous détaillons différentes optimisations possibles en OpenCL, pour finalement définir une stratégie d'exploration pertinente pour l'accélération d'algorithmes sur FPGA. Appliquée sur différents cas d'étude, de la reconstruction tomographique à la modélisation d'un brouillage aéroporté radar, nous évaluons notre méthodologie suivant trois principaux critères de performance : le temps de développement, le temps d'exécution, et l'efficacité énergétique. / As the semiconductor industry faces major challenges in sustaining its growth, new High-Level Synthesis tools are repositioning FPGAs as a leading technology for algorithm acceleration in the face of CPU and GPU-based clusters. But as it stands, for a software engineer, these tools do not guarantee, without expertise of the underlying hardware, that these technologies will be harnessed to their full potential. This can be a game breaker for their democratization. From this observation, we propose a methodology for algorithm acceleration on FPGAs. After presenting a high-level model of this architecture, we detail possible optimizations in OpenCL, and finally define a relevant exploration strategy for accelerating algorithms on FPGA. Applied to different case studies, from tomographic reconstruction to the modelling of an airborne radar jammer, we evaluate our methodology according to three main performance criteria: development time, execution time, and energy efficiency.

No description available.

No description available.

No description available.

De senaste årens utveckling av generativ AI har fått mycket uppmärksamhet i etablerade medier och generellt i samhället, vilket bland annat har handlat om bildgenereringsverktyg. Något som det saknas tidigare forskning kring är hur text-tillbild AI kan användas inom webbdesign. Den befintliga forskningen som finns på området tyder på att det behövs skapas nya designprocesser för människa-AI interaktion. I denna uppsats undersöks hur designers kan ta användning av AI och om det i sig kan ses som en samarbetspartner eller endast ett verktyg. Uppsatsens frågeställning är: På vilka sätt kan AI i en samskapande process implementeras inom området webbdesign? Arbetetgrundar sig i tidigare forskning inom AI som hjälpmedel i innovation, implementationen av generativ AI i en designprocess, AI inom UX-design samt kommunikation i en samskapande process. Dessutom utgår även studien från tidigare forskning inom MidJourney i en kreativt skapande process, samt de teoretiska perspektiven design thinking och människocentrerad design. Metoderna som används är ett användartest av MidJourney och intervjuer med professionella webbdesigners som deltagare. En majoritet av deltagarna ansåg att MidJourney hade svårt att följa designkriterier som främst handlade om användarvänlighet men var överlag fortsatt positiva till verktygets förmåga. Alla deltagare uttryckte att de hade kunnat fortsätta använda MidJourney i framtida arbeten och nämnde att den kunde implementeras i en designprocess när det kommer till att underlätta och effektivisera kommunikation med klienten för visuella idéer, som ersättning för traditionella inspirationskällor, men främst att den kan vara ett bra verktyg vid idégenerering. Samarbetet med verktyget hade enligt deltagarna kunnat fungera bättre om kommunikation som användes vid användandet var mer mänsklig. Slutsatsen blir att en AI i nuläget kan implementeras i flera steg i en designprocess, som en stöttning i kreativa processer och för att effektivisera mycket arbete. Det framkommer även att AI:n inte kan ses som en samarbetspartner utan endast som ett verktyg som kan användas inom flera områden i en designprocess. Men i framtiden kommer AI i större utsträckning upplevas som en samarbetspartner, genom utbilda designerns mer inom AI och utveckla kommunikation till mer människolik. Idag förblir designers roll viktig i och med AI:s svårigheter att förstå grundläggande design thinking och människocentrerad design. / In recent years development of generative AI has received a lot of attention in established media and generally in society, which among other things has been about image generation tools. One area where there is a lack of research in, is how text-to-image AI can be used in web design. Existing research suggests that new design processes for human-AI interaction needs to be created. This essay explores how designers can make use of AI and whether it can be seen as a collaborative partner or a tool. The research question for the essay is: In what ways can AI be implemented in a co-creation process in the web design field? The work is based on previous research on AI as a tool in innovation, implementation of generative AI in a design process, AI in UX design, and communication in a co-creation process. Additionally, the study is also based on previous research on MidJourney in a creative process and the theoretical perspectives of design thinking and human-centered design. The methods that are being used are interviews and a user test with professional web designers as participants. The majority of the participants felt that MidJourney had difficulty following design criterias that primarily concerned usability, but were still positive about the tool's ability. All participants expressed that they could continue to use MidJourney in future works and mentioned that it could be implemented in a design process to facilitate and improve efficiency in communication with the client for visual ideas, as a replacement for traditional inspiration sources, but primarily as a good tool for idea generation. According to the participants, the collaboration with the tool could have worked better if the communication used during its use was more human-like. The conclusion is that AI currently can be implemented in several steps in a design process, as a support in creative processes and to improve efficiency in a lot of work. It also emerges that AI cannot be seen as a collaborative partner, only as a tool that can be used in several areas. But in thefuture, AI will increasingly be perceived as a collaborative partner, by educating designers more in AI and developing communication to be more human-like. However, the designer's role remains important today due to AI:s difficulties in understanding basic design thinking and human-centered design.

The promise of smart cities to deliver new urban efficiencies and optimizations for sustainability is increasingly being questioned for its anthropocentric, universal, and top-down perspectives. Framingcities as computers has been critiqued for its limiting understanding of cities, as well as its lack of dealing with the complexities of real messy cities, with diverse knowledge and lived experiences. However, smart technologies have also been highlighted as having the potential to help us better understand more-than-human perspectives and to reconnect us to the world around us. Situated in thefield of design for social innovation, this thesis contributes to the emerging body of work that is exploring how digital urban environments can include local knowledge and more-than-human perspectives. In a co-design process with the urban agriculture community of Sjöbergen in the city of Gothenburg in Sweden, this thesis explores how local knowledge and values about- and in urban nature can help us think differently about the future of sustainable smart city concepts. With a design process guided by research through design and co-design, this thesis is imagining other smart city narratives that go away from the current top-down and universal perspectives and instead are inspired by values of Sjöbergen of reuse, maintenance, collectivity, and knowledge sharing. The design contribution of this work is a design proposal of a smart city service that reuses old smartphones of citizens into smart city technologies for individual and situated purposes. The design proposal aims to show an alternative view of smart cities grounded in local values and more-than-human perspectives.