Fourth Generation (4G) cellular technology Third Generation Partnership Project (3GPP) Long Term Evolution (LTE) offers high data rate capabilities to mobile users; and, operators are trying to deliver a true mobile broadband experience over LTE networks. Mobile TV and Video on Demand (VoD) are expected to be the main revenue generators in the near future [36] and efficient video streaming over wireless is the key to enabling this. 3GPP recommends the use of H.264 baseline profiles for all video based services in Third Generation (3G) Universal Mobile Telecommunication System (UMTS) networks. However, LTE networks need to support mobile devices with different display resolution requirements like small resolution mobile phones and high resolution laptops. Scalable Video Coding (SVC) is required to achieve this goal. Feasibility study of SVC for LTE is one of the main agenda of 3GPP Release10. SVC enhances H.264 with a set of new profiles and encoding tools that may be used to produce scalable bit streams. Efficient adaptation methods for SVC video transmission over LTE networks are proposed in this thesis. Advantages of SVC over H.264 are analyzed using real time use cases of mobile video streaming. Further, we study the cross layer adaptation and scheduling schemes for delivering SVC video streams most efficiently to the users in LTE networks in unicast and multicast transmissions. We propose SVC based video streaming scheme for unicast and multicast transmissions in the downlink direction, with dynamic adaptations and a scheduling scheme based on channel quality information from users. Simulation results indicate improved video quality for more number of users in the coverage area and efficient spectrum usage with the proposed methods.

We propose a new ensemble algorithm: the meta-boosting algorithm. This algorithm enables the original Adaboost algorithm to improve the decisions made by different WeakLearners utilizing the meta-learning approach. Better accuracy results are achieved since this algorithm reduces both bias and variance. However, higher accuracy also brings higher computational complexity, especially on big data. We then propose the parallelized meta-boosting algorithm: Parallelized-Meta-Learning (PML) using the MapReduce programming paradigm on Hadoop. The experimental results on the Amazon EC2 cloud computing infrastructure show that PML reduces the computation complexity enormously while retaining lower error rates than the results on a single computer. As we know MapReduce has its inherent weakness that it cannot directly support iterations in an algorithm, our approach is a win-win method, since it not only overcomes this weakness, but also secures good accuracy performance. The comparison between this approach and a contemporary algorithm AdaBoost.PL is also performed.

In this work the development of a Self-Powered System-On-Chip is explored by examining two components of process development in different perspectives. On one side, an energy component is approached from a biochemical standpoint where a Microbial Fuel Cell (MFC) is built with standard microfabrication techniques, displaying a novel electrode based on Carbon Nanotubes (CNTs). The fabrication process involves the formation of a micrometric chamber that hosts an enhanced CNT-based anode. Preliminary results are promising, showing a high current density (113.6mA/m2) compared with other similar cells. Nevertheless many improvements can be done to the main design and further characterization of the anode will give a more complete understanding and bring the device closer to a practical implementation. On a second point of view, nano-patterning through silicon nitride spacer width control is developed, aimed at producing alternative sub-100nm device fabrication with the potential of further scaling thanks to nanowire based structures. These nanostructures are formed from a nano-pattern template, by using a bottom-up fabrication scheme. Uniformity and scalability of the process are demonstrated and its potential described. An estimated area of 0.120μm2 for a 6T-SRAM (Static Random Access Memory) bitcell (6 devices) can be achieved. In summary, by using a novel sustainable energy component and scalable nano-patterning for logic and computing module, this work has successfully collected the essential base knowledge and joined two different elements that synergistically will contribute for the future implementation of a Self-Powered System-on-Chip.

Photovoltaics provide a feasible route to fulfilling the substantial increase in demand for energy worldwide. Solution processable organic photovoltaics (OPVs) have attracted attention in the last decade because of the promise of low-cost manufacturing of sufficiently efficient devices at high throughput on large-area rigid or flexible substrates with potentially low energy and carbon footprints. In OPVs, the photoactive layer is made of a bulk heterojunction (BHJ) layer and is typically composed of a blend of an electron-donating (D) and an electron-accepting (A) materials which phase separate at the nanoscale and form a heterojunction at the D-A interface that plays a crucial role in the generation of charges. Despite the tremendous progress that has been made in increasing the efficiency of organic photovoltaics over the last few years, with power conversion efficiency increasing from 8% to 13% over the duration of this PhD dissertation, there have been numerous debates on the mechanisms of formation of the crucial BHJ layer and few clues about how to successfully transfer these lessons to scalable processes. This stems in large part from a lack of understanding of how BHJ layers form from solution. This lack of understanding makes it challenging to design BHJs and to control their formation in laboratory-based processes, such as spin-coating, let alone their successful transfer to scalable processes required for the manufacturing of organic solar cells. Consequently, the OPV community has in recent years sought out to better understand the key characteristics of state of the art lab-based organic solar cells and made efforts to shed light on how the BHJ forms in laboratory-based processes as well as in scalable processes. We take the view that understanding the formation of the solution-processed bulk heterojunction (BHJ) photoactive layer, where crucial photovoltaic processes take place, is the one of the most crucial steps to developing strategies towards the implementation of organic solar cells with high efficiency and manufacturability. In this dissertation, we investigate the mechanism of the BHJ layer formation during solution processing from common lab-based processes, such as spin-coating, with the aim of understanding the roles of materials, formulations and processing conditions and subsequently using this insight to enable the scalable manufacturing of high efficiency organic solar cells by such methods as wire-bar coating and blade-coating. To do so, we have developed state-of-the-art in situ diagnostics techniques to provide us with insight into the thin film formation process. As a first step, we have developed a modified spin-coater which allows us to perform in situ UV-visible absorption measurements during spin coating and provides key insight into the formation and evolution of polymer aggregates in solution and during the transformation to the solid state. Using this method, we have investigated the formation of organic BHJs made of a blend of poly (3-hexylthiophene) (P3HT) and fullerene, reference materials in the organic solar cell field. We show that process kinetics directly influence the microstructure and morphology of the bulk heterojunction, highlighting the value of in situ measurements. We have investigated the influence of crystallization dynamics of a wide-range of small-molecule donors and their solidification pathways on the processing routes needed for attaining high-performance solar cells. The study revealed the reason behind the need of empirically-adopted processing strategies such as solvent additives or alternatively thermal or solvent vapor annealing for achieving optimal performance. The study has provided a new perspective to materials design linking the need for solvent additives or annealing to the ease of crystallization of small-molecule donors and the presence or absence of transient phases before crystallization. From there, we have extended our investigation to small-molecule (p-DTS (FBTTh2)2) fullerene blend solar cells, where we have revealed new insight into the crucial role of solvent additives. Our work has also touched upon modern polymers, such as PBDTTPD, where we have found the choice of additives impacts the formation mechanism of the BHJ. Finally, we have performed a comparative study of the BHJ film formation dynamics during spin coating versus wire-bar coating of p-DTS(FBTTh2)2: fullerene blends that has helped in curbing the performance gap between lab-based and scalable techniques. This was done by implementing a new apparatus that combines the benefits of rapid thin film drying common to spin coating with scalability of wire-bar coating. Using the new apparatus, we successfully attain similar performance of solar cell devices to the ones fabricated by spin coating with dramatically reduced material waste.

Next-generation soft robots, wearable health monitoring devices, and human-machine interfaces require electronic systems that can maintain their performance under deformations. Thus, researchers have been developing materials and methods to enable high-performance soft electronic systems in diverse applications. While a variety of solutions have been presented, development of stretchable materials with a combination of high stretchability, electrical conductivity, cyclic stability, and manufacturability is still an open challenge. Throughout this dissertation, gallium-based<br>liquid metal alloy is used as the conductive material, leveraging its high conductivity and intrinsic stretchability for maintained performance under deformations. This dissertation presents both new liquid metal-based conductive materials and scalable manufacturing methods for the development of a diverse range of flexible and stretchable electronic circuits. First, a laser sintering method was developed to coalesce liquid metal micro/nanoparticles into soft, conductive structures enabled by oxide rupturing. The fast, non-contact, and maskless laser sintering technique, in combination with large-area spray-printing deposition, and high-throughput emulsion processing, provided a methodology to create different physical manifestations of liquid metal-based soft, stretchable, and reconfigurable electronics. Second, a liquid metal-based biphasic material was created using a thermal processing technique, yielding a printable, mechanically stable, and extremely stretchable conductor. This material’s compatibility with existing scalable manufacturing methods, robust interfaces with off-the-shelf electronic components, and electrical/mechanical cyclic stability enabled direct conversion of established circuit board assemblies to stretchable forms. The work presented in this dissertation paves the way for future mass-manufacturing of<br>soft, stretchable circuits for direct integration into smart garments or soft robots. <br>

The dissertation focuses on detecting contextual outliers from heterogeneous data sources. Modern sensor-based applications such as Internet of Things (IoT) applications and autonomous vehicles are generating a huge amount of heterogeneous data including not only the structured multi-variate data points, but also other complex types of data such as time-stamped sequence data and image data. Detecting outliers from such data sources is critical to diagnose and fix malfunctioning systems, prevent cyber attacks, and save human lives. The outlier detection techniques in the literature typically are unsupervised algorithms with a pre-defined logic, such as, to leverage the probability density at each point to detect outliers. Our analysis of the modern applications reveals that this rigid probability density-based methodology has severe drawbacks. That is, low probability density objects are not necessarily outliers, while the objects with relatively high probability densities might in fact be abnormal. In many cases, the determination of the outlierness of an object has to take the context in which this object occurs into consideration. Within this scope, my dissertation focuses on four research innovations, namely techniques and system for scalable contextual outlier detection from multi-dimensional data points, contextual outlier pattern detection from sequence data, contextual outlier image detection from image data sets, and lastly an integrative end-to-end outlier detection system capable of doing automatic outlier detection, outlier summarization and outlier explanation. 1. Scalable Contextual Outlier Detection from Multi-dimensional Data. Mining contextual outliers from big datasets is a computational expensive process because of the complex recursive kNN search used to define the context of each point. In this research, leveraging the power of distributed compute clusters, we design distributed contextual outlier detection strategies that optimize the key factors determining the efficiency of local outlier detection, namely, to localize the kNN search while still ensuring the load balancing. 2. Contextual Outlier Detection from Sequence Data. For big sequence data, such as messages exchanged between devices and servers and log files measuring complex system behaviors over time, outliers typically occur as a subsequence of symbolic values (or sequential pattern), in which each individual value itself may be completely normal. However, existing sequential pattern mining semantics tend to mis-classify outlier patterns as typical patterns due to ignoring the context in which the pattern occurs. In this dissertation, we present new context-aware pattern mining semantics and then design efficient mining strategies to support these new semantics. In addition, methodologies that continuously extract these outlier patterns from sequence streams are also developed. 3. Contextual Outlier Detection from Image Data. An image classification system not only needs to accurately classify objects from target classes, but also should safely reject unknown objects that belong to classes not present in the training data. Here, the training data defines the context of the classifier and unknown objects then correspond to contextual image outliers. Although the existing Convolutional Neural Network (CNN) achieves high accuracy when classifying known objects, the sum operation on multiple features produced by the convolutional layers causes an unknown object being classified to a target class with high confidence even if it matches some key features of a target class only by chance. In this research, we design an Unknown-aware Deep Neural Network (UDN for short) to detect contextual image outliers. The key idea of UDN is to enhance existing Convolutional Neural Network (CNN) to support a product operation that models the product relationship among the features produced by convolutional layers. This way, missing a single key feature of a target class will greatly reduce the probability of assigning an object to this class. To further improve the performance of our UDN at detecting contextual outliers, we propose an information-theoretic regularization strategy that incorporates the objective of rejecting unknowns into the learning process of UDN. 4. An End-to-end Integrated Outlier Detection System. Although numerous detection algorithms proposed in the literature, there is no one approach that brings the wealth of these alternate algorithms to bear in an integrated infrastructure to support versatile outlier discovery. In this work, we design the first end-to-end outlier detection service that integrates outlier-related services including automatic outlier detection, outlier summarization and explanation, human guided outlier detector refinement within one integrated outlier discovery paradigm. Experimental studies including performance evaluation and user studies conducted on benchmark outlier detection datasets and real world datasets including Geolocation, Lighting, MNIST, CIFAR and the Log file datasets confirm both the effectiveness and efficiency of the proposed approaches and systems.

This thesis was performed at the company ShapeUp Club located in Stockholm, Sweden. ShapeUp Club offers a digital calorie counter service for the web, iOS and Android with data synchronization across the platforms. ShapeUp Club wants to provide their users with the option to synchronize data between ShapeUp Club and external health services. The objective for this thesis has been to develop an extension to ShapeUp Clubs current backend platform where new external health services can be plugged-in quickly and scalable. External partner APIs will be examined and implemented in the system to validate the functionality of the system. The amount of code needed to plug-in a service should be as minimal as possible for a developer to quickly add another service. To allow for scalability the platform also needs to adapt logic for how often users should be allowed to poll for data from their connected services, to minimize the database load for all parts. To handle these demands, an extension to ShapeUp Club’s current backend solution was built using the Django framework for Python. By providing a generic base class that new services inherit from, the amount of code necessary for implementing a new service is reduced to methods for API- requests, authorization and serialization of data. To reduce the number of redundant poll requests, users are placed into groups. Each group is a cluster of users with similar frequency of updates. Django’s cache framework is used to handle the concurrency of the sync tasks, which locks a user from syncing the same partner in parallel. / Detta examensarbete har utförts hos företaget ShapeUp Club i Stockholm. ShapeUp Club erbjuder en digital kaloriräknare för webben, iOS och Android med synkronisering av data mellan dessa plattformar. ShapeUp Club vill kunna erbjuda sina kunder möjligheten att synkronisera data mellan ShapeUp Club och andra externa hälsotjänster. Målet med detta projekt har varit att implementera en ny tjänst till ShapeUp Clubs nuvarande backend-lösning där externa hälsotjänster snabbt och skalbart kan implementeras. Externa hälso-API:er har utvärderats och implementerats i samband med utvecklingen av den nya backendtjänsten, för att validera dess funktionalitet. Mängden kod som behövs för att implementera en hälsotjänst bör vara så minimal som möjligt för att utvecklare snabbt ska kunna lägga till ytterligare tjänster. För att systemet ska vara skalbart måste logik finnas för hur ofta användare ska tillåtas att fråga efter data mot de tjänster de har valt att synkronisera mot. För att tillfredställa dessa behov har en utökning av ShapeUp Clubs nuvarande backend-lösning byggts med ramverket Django för Python. Genom att ha en större, generisk klass som nya implementeringar ärver från så har mängden nödvändig kod för varje hälsotjänst-implementering minskats till metoder för API-anrop, autentisering och serialisering av data. För att minska antalet “onödiga” poll-anrop så placerar vi användare i olika grupper beroende på om deras poll-anrop frekvent återvänder utan någon ny information. De olika grupperna bestämmer sedan hur länge användarna måste vänta innan de tillåts göra nya poll-anrop.

In this thesis context refers to information about the environment (the user or entity's surroundings) that can influence and determine the behavior of a computing system. Context-awareness means that the computer can adapt to the situation in which it is working. Context is a key issue in mobile computing, especially with handheld devices (such as PDAs and mobile phones), due to the fact that they can be used while on the move; hence the environment around them can change. The environment of a static device may also change and require the device to adapt. Applications and systems that exploit context by both sensing and reacting to their environment are called context-aware applications. Devices that are context-aware are able to perceive stimuli and react accordingly, with minimal interaction with the user. Providing context-aware services to users of mobile devices via context-aware applications is becoming an important and significant factor in the market and is a developing industry. In this thesis we analyze and develop an application that exploits context to provide a service that improves the interaction between humans and a computer. The thesis begins with a study of what types of sensors are available to provide information about the device's context. This is followed by the design of an appropriate way of using the selected sensor (ecompass) to provide a means of adapting a service to the user's and device's context. The focus is every day activities of a student at a university - specifically finding the location of a meeting room for a seminar; however, similar scenarios exist for other types of users. After determining that it was feasible to add a e-compass as a sensor to an existing personal digital assistant and to provide a map to the mobile user, the focus of the thesis shifted to an examination of the performance of the adaptation of the map as the user moved the device. Initially it required excessive time to render the map on the device, thus as the user moved the device the map was not updated quickly enough for the user to know their correct orientation with respect to the map. Therefore the thesis project examined how this performance could be improved sufficiently that the rendering would keep up with the change in orientation of the device. This investigation lead to a shift from server based rendering of the map as an image, followed by the transfer of the image to the device for display; to a sending a scalable vector graphics version of the map to the device for local rendering. While initially this was expected to be much faster than transferring an image for an actual map of the building where this work was taking place the local rendering was actually slower. This subsequently lead to server based pruning of the irrelevant details from the map, then a transfer of the relevant portion of the map to the device, followed by local rendering. / I den här avhandlingen hänvisar 'context' till information om miljön (i användarens eller enhetens omgivning) som kan bestämma och påverka beteendet hos ett datorsystem. Contex-awareness innebär att datorn kan anpassa sig till den situation där den arbetar. Context är en central fråga för mobila enheter, speciellt för handhållna enheter (t.ex. handdatorer och mobiltelefoner), på grund av att de kan användas på resande fot där omgivningen hela tiden förändras. Omgivningen för en statisk enhet kan också förändras och kräver att enheten kan anpassa sig. Applikationer och system som utnyttjar context genom att både känna av och reagera på sin omgivning kallas context-aware applications. Enheter som är kontextmedvetna kan uppfatta stimuli och reagera på den med minimal användarinteraktion. Att tillhandahålla kontextmedvetna tjänster till användare av mobila enheter via kontextmedvetna applikationer blir en allt viktigare och betydelsefullare faktor på marknaden och är en växande industri. I den här avhandlingen analyserar och utvecklar vi ett program som utnyttjar kontext för att tillhandahålla en tjänst som förbättrar samspelet mellan människa och dator. Avhandlingen inleds med en undersökning av vilka typer av sensorer som finns tillgängliga för att tillhandahålla information om enhetens kontext. Detta följs av en design för att på lämpligaste sätt använda den valda sensorn (e-kompass) för att tillhandahålla ett sätt att anpassa en tjänst till användaren och enhetens kontext. Fokus är vardagsaktivitieter för en student vid ett universitet - särskilt att hitta till ett konferensrum för ett seminarium, liknande scenarier finns även för andra typer av användare. Efter att ha fastställt att det var möjligt att koppla en sensor, i form av en e-kompass, till en befintlig personal digital assistant samt att visa en karta för användaren, flyttades fokus för avhandlingen till en undersökning om tjänstens prestanda när användaren flyttade enheten. Initialt krävde enheten väldigt lång tid att rendera kartan och när enheten flyttades uppdaterades kartan inte tillräckligt snabbt för att användaren skulle veta sin riktning i relation till kartan. Därför undersöktes hur prestandan kunde förbättras så att enheten skulle kunna hänga med bättre när enhetens riktning förändrades. Undersökningen ledde till att istället för att rendera en bild på servern och sedan skicka till enheten, skapa en vektorbaserad bild på servern, skicka till enheten och rendera lokalt. Även om detta initialt förväntades vara mycket snabbare än att överföra en bild av en verklig karta var den lokala renderingen faktiskt ännu långsammare. Detta ledde till en serverbaserad gallring av ovidkommande kartdetaljer samt beskärning innan kartan fördes över till enheten och renderades lokalt.

Due to their suitability for highly parallel and pipelined computation, field programmable gate arrays (FPGAs) and general-purpose graphics processing units (GPGPUs) have emerged as top contenders for hardware acceleration of high-performance computing applications. FPGAs are highly specialized devices that can be customized to a specific application, whereas GPGPUs are made of a fixed array of multiprocessors with a rigid architectural model. To alleviate this rigidity as well as to combine some other benefits of the two platforms, it is desirable to explore the implementation of a flexible GPGPU (soft GPGPU) using the reconfigurable fabric found in an FPGA. This thesis describes an aggressive effort to test and validate a prototype GPGPU design targeted to a Virtex-6 FPGA. Individual design stages are tested and integrated together using manually-generated RTL testbenches and logic simulation tools. The soft GPGPU design is validated by benchmarking the platform against five standard CUDA benchmarks. The platform is fully CUDA-compatible and supports direct execution of CUDA compiled binaries. Platform scalability is validated by varying the number of processing cores as well as multiprocessors, and evaluating their effects on area and performance. Experimental results show as average speedup of 25x for a 32 core soft GPGPU configuration over a fully optimized MicroBlaze soft microprocessor, accentuating benefits of the thread-based execution model of GPUs and their ability to perform complex control flow operations in hardware. The testing and validation of the designed soft GPGPU system serves as a prerequisite for rapid design exploration of the platform in the future.

As silicon processes scale, system-on-chips (SoCs) will require numerous hardware monitors that perform assessment of physical characteristics that change during the operation of a device. To address the need for high-speed and coordinated transport of monitor data in a SoC, we develop a new interconnection network for monitors - the monitor network on chip (MNoC). Data collected from the monitors via MNoC is collated by a monitor executive processor (MEP) that controls the operation of the SoC in response to monitor data. In this thesis, we developed the architecture of MNoC and the infrastructure to evaluate its performance and overhead for various network parameters. A system level architectural simulation can then be performed to ensure that the latency and bandwidth provided by MNoC are sufficient to allow the MEP to react in a timely fashion. This typically translates to a system level benefit that can be assessed using architectural simulation. We demonstrate in this thesis, the employment of MNoC for two specific monitoring systems that involve thermal and delay monitors. Results show that MNoC facilitates employment of a thermal-aware dynamic frequency scaling scheme in a multicore processor resulting in improved performance. It also facilitates power and performance savings in a delay -monitored multicore system by enabling a better than worst case voltage and frequency settings for the processor.