The recent proliferation of sensors in inhospitable environments such as disaster or battle zones has not been matched by in situ data processing capabilities due to a lack of computing infrastructure in the field. We envision a solution based on small, low-altitude unmanned aerial vehicles (UAVs) that can deploy elastically-scalable computing infrastructure anywhere, at any time. This airborne compute cloud—essentially, micro-data centers hosted on UAVs—would communicate with terrestrial assets over a bandwidth-constrained wireless network with variable, unpredictable link qualities. Achieving high performance over this ground-to-air mobile radio channel thus requires making full and efficient use of every single transmission opportunity. To this end, this dissertation presents two system primitives that improve throughput and reduce network overhead by using recent distributed coding methods to exploit natural properties of the airborne environment (i.e., antenna beam diversity and anomaly sparsity). We first built and deployed an UAV wireless networking testbed and used it to characterize the ground-to-UAV wireless channel. Our flight experiments revealed that antenna beam diversity from using multiple SISO radios boosts reception range and aggregate throughput. This observation led us to develop our first primitive: ground-to-UAV bulk data transport. We designed and implemented FlowCode, a reliable link layer for uplink data transport that uses network coding to harness antenna beam diversity gains. Via flight experiments, we show that FlowCode can boost reception range and TCP throughput as much as 4.5-fold. Our second primitive permits low-overhead cloud status monitoring. We designed CloudSense, a network switch that compresses cloud status streams in-network via compressive sensing. CloudSense is particularly useful for anomaly detection tasks requiring global relative comparisons (e.g., MapReduce straggler detection) and can achieve up to 16.3-fold compression as well as early detection of the worst anomalies. Our efforts have also shed light on the close relationship between network coding and compressive sensing. Thus, we offer FlowCode and CloudSense not only as first steps toward the airborne compute cloud, but also as exemplars of two classes of applications—approximation intolerant and tolerant—to which network coding and compressive sensing should be judiciously and selectively applied. / Engineering and Applied Sciences

Den finansiella utbildningen präglas av klassisk finansteori som förutsätter att den finansiella marknaden prissätts rationellt. Det finns dock ett gap mellan klassisk finansteori och verklighet. Syftet har därför varit att se hur en investerare genom ökad medvetenhet om dessa anomalier kan förbättra aktiemarknadsanalys och investeringsbeslut. Studien har genomförts med ett kvalitativt tillvägagångssätt och baserats på en litteraturstudie som kompletterats med intervjuer. Under studien har en bild av investeraren som begränsat rationell framträtt i linje med de teorier som har redovisats. Där flockbeteende vuxit fram som det mest påtagliga stödet för att den klassiska finansteorin inte är att likställa med marknadens dynamiska verklighet. I studien har teorierna inom behavioral finance tematiserats och redogjorts för mot bakgrund av det empiriska underlaget. Gemensamt är att investerare tenderar att vara begränsat rationella. Psykologin är ständigt närvarande i marknaden och påverkar investerare i deras beslutsfattande i större utsträckning än vad klassisk finansteori ger utrymme för. Detta är ett av de främsta skälen till varför behavioral finance och dess teorier borde bli ett komplement till den klassiska finansteorin. Slutsatsen är att det finns möjligheter för investerare att förbättra aktiemarknadsanalys och investeringsbeslut genom att ta teorierna inom behavioral finance i beaktning. / The financial education is characterized by classical financial theory that assumes that the fi-nancial market is priced rationally. However, there is a gap between classic finance theory and reality. The aim has been to see how an investor through increased awareness of these anomalies can improve stock market analysis and investment decisions. The study was conducted with a qualitative approach and was based on a literature review supplemented by interviews. During the study, proofs of semi-rational investors have emerged in line with the theories of behavioral finance. Herd behavior has emerged as the most tangible proof that the classical financial theory is not comparable to the dynamic reality of the market. In the study, theories of behavioral finance has been thematised and explained in the light of the empirical basis. In common for those theories is that investors tend to be semi-rational. The psychology is always present in the market and affects investors in their decision making to a greater extent than classic finance theories allow. It is one of the main reasons why it should be implemented as a complement to the traditional financial theories. The conclusion is that there is potential for investors to improve stock market analysis and investment decisions by taking theories of behavioral finance into consideration.

With the manufacture of nano-scale features in the last ten years, it is possible to do optical experiments on features as small as a tenth/hundredth wavelength. It turns out that the experimental data cannot be explained by classical diffraction theories. Thus, it is necessary to develop new methods or use existing approaches which are effective in other fields, to solve problems in photonics. We use finite difference time domain (FDTD), to study transmission properties of sub-wavelength slits in a metallic film. By doing simulations on periodic and single slits, we confirm that the TE mode has a cutoff while a TM mode always has a propagating mode in the small apertures. Then we find that the transmittance is minimum when the array period is equal to the wavelength of surface plasmon polariton (SPP) at normal incidence. In fact, the SPP-like waves exist in both periodic and isolated slits, and they help the transmittance of small apertures. In order to establish the role of SPP in the transmission mechanism, it is necessary to single out each mode from the total fields. We developed Bloch mode method (BMM) to calculate the amplitudes of the lowest N orders, and the amplitudes tell us which one is dominant (not including the guided mode) at high and low transmission. BMM converges very fast and it is more accurate than FDTD since it does not suffer from numerical dispersion. Both methods can resolve the Wood anomaly and SPP anomaly; however, FDTD converges very slowly at the SPP resonance and oscillates around the value obtained through BMM at the Wood anomaly. BMM is not sensitive to material types, incident angles, and anomalies; it will be a useful tool to investigate similar problems.

Bakgrund: Effektiva marknadshypotesen innebär att det inte går att utnyttja systematiska avvikelser på marknaden. Trots det finns det etablerade investeringsstrategier som investerare använder sig av för att generera överavkastning. Syfte: Syftet med studien är att undersöka huruvida det går att generar överavkastning genom att investera i företag som uppvisar en låg eller hög EV/EBITDA-multipel. Variablerna bransch och risk kommer även att undersökas med utgångspunkt från den eventuella förekomsten av en investeringsstrategi som genererar överavkastning. Genomförande: Teori inom området har byggt upp en grundförståelse för problemet, empiri har sedan hämtats från de olika företagen för att få fram EV/EBITDA-multiplar till de ingående portföljerna i studien. Aktiekurser har även inhämtats för att bygga grunden till empirin. Resultatet har sedan jämförts med OMXSPI samt med den riskjusterade avkastningen. Resultat: Av resultatet framkommer att det går att utnyttja en investeringsstrategi där investering görs i låga EV/EBITDA-multiplar. Effekten är tydlig på hela Stockholmsbörsen, samt i två av de undersökta branscherna, sällanköpsvaror och tillverkningsindustrin. Det spelar ingen roll vilken riskpreferens investeraren har då portföljer med låga multiplar genererar högst avkastning och även innehåller lägst betavärde. / Background: The efficient market hypothesis alleges that an investor cannot systematically earn excess return. However there are established investment strategies that are being used in the stock market to obtain this excess return Aim: The aim of the thesis is to examine if it is possible to earn excess returns by investing in companies that indicate a low or high EV/EBITDA multiple. The variables industry and risk will also be examined with the possible presence of an investment strategy where excess return can be obtained as a base. Completion: Theory within the field has built an understanding of the problem, empirics have then been gathered to obtain EV/EBITDA multiples and stock prices to perform the study. Portfolios of high and low multiples have been composed to analyze the result. The result has been compared to the development of the OMXSPI index and to the risk adjusted return for the portfolios. Result: The result shows that an investment strategy where investments are made in a low EV/EBITDA-multiple can be used to earn excess return on the Swedish stock market. The effect is most present at the Stockholm stock exchange and in the manufacturing industry and industry for durable goods. The investors risk aversion does not affect the decision, thus low multiples generate the highest return with the lowest beta value.

To United Minds, large-scale surveys are an important offering to clients, not least the public opinion poll Väljarbarometern. A risk associated with surveys is satisficing – sub-optimal response behaviour impairing the possibility of correctly describing the sampled population through its results. The purpose of this study is to – through the use of multivariate outlier detection methods - identify those observations assumed to be non-representative of the population. The possibility of categorizing responses generated through satisficing as outliers is investigated. With regards to the character of the Väljarbarometern dataset, three existing algorithms are adapted to detect these outliers. Also, a number of randomly generated observations are added to the data, by all algorithms correctly labelled as outliers. The resulting anomaly scores generated by each algorithm are compared, concluding the Otey algorithm as the most effective for the purpose, above all since it takes into account correlation between variables. A plausible cut-off value for outliers and separation between non-representative and representative outliers are discussed. The resulting recommendation is to handle observations labelled as outliers through respondent follow-up or if not possible, through downweighting, inversely proportional to the anomaly scores.

This paper describes various network monitoring technologies and anomaly detection methods. NetFlow were chosen for anomaly detection system being developed. Anomalies are detected using a deviation value. After evaluating quality of developed system, new enhancements were suggested and implemented. Flow data distribution was suggested, to achieve more precise NetFlow data representation, enabling a more precise network monitoring information usage for anomaly detection. Arithmetic average calculations were replaced with more flexible Exponential Weighted Moving Average algorithm. Deviation weight was introduced to reduce false alarms. Results from experiment with real life data showed that proposed changes increased precision of NetFlow based anomaly detection system.

The continued increase in the size and complexity of modern computer systems has led to a commensurate increase in the size of their logs. System logs are an invaluable resource to systems administrators during fault resolution. Fault resolution is a time-consuming and knowledge intensive process. A lot of the time spent in fault resolution is spent sifting through large volumes of information, which includes event logs, to find the root cause of the problem. Therefore, the ability to analyze log files automatically and accurately will lead to significant savings in the time and cost of downtime events for any organization. The automatic analysis and search of system logs for fault symptoms, otherwise called alerts, is the primary motivation for the work carried out in this thesis. The proposed log alert detection scheme is a hybrid framework, which incorporates anomaly detection and signature generation to accomplish its goal. Unlike previous work, minimum apriori knowledge of the system being analyzed is assumed. This assumption enhances the platform portability of the framework. The anomaly detection component works in a bottom-up manner on the contents of historical system log data to detect regions of the log, which contain anomalous (alert) behaviour. The identified anomalous regions are then passed to the signature generation component, which mines them for patterns. Consequently, future occurrences of the underlying alert in the anomalous log region, can be detected on a production system using the discovered pattern. The combination of anomaly detection and signature generation, which is novel when compared to previous work, ensures that a framework which is accurate while still being able to detect new and unknown alerts is attained. Evaluations of the framework involved testing it on log data for High Performance Cluster (HPC), distributed and cloud systems. These systems provide a good range for the types of computer systems used in the real world today. The results indicate that the system that can generate signatures for detecting alerts, which can achieve a Recall rate of approximately 83% and a false positive rate of approximately 0%, on average.

At the present time it is unknown how the U(1)A anomaly of Quantum Chromodynamics behaves at high temperatures. We therefore want to look for thermal changes of the effects of the anomaly. For example, by studying the properties of the η' meson at high temperatures it would be possible to deduce important information on the axial anomaly, thanks to the deep connection between them. In this thesis the width of the η' as a function of the temperature is studied in the framework of large-Nc Chiral Perturbation Theory, at next-to-leading order, and in the corresponding Resonance Chiral Theory. We calculate the width increase due to scattering with particles from the heat bath, which we assume to consist of a pion gas. We compare the results obtained in both frameworks and as expected we find a smaller, but still consistent width increase when the more realistic resonance exchange is taken into account. The results suggest that the in-medium width of the η' may increase up to ΔΓ<img src="http://www.diva-portal.org/cgi-bin/mimetex.cgi?%5Capprox" /> 10 MeV at a temperature of T<img src="http://www.diva-portal.org/cgi-bin/mimetex.cgi?%5Capprox" /> 120 MeV. We find therefore a width increase of considerable size, comparable to the inverse lifetime of the fireball created in relativistic heavy-ion collisions. In other words, our results suggest that it may be possible to study experimentally how the properties of the η' change at high temperatures.

We consider the problem of extracting a specific feature from a noisy signal generated by a multi-channels Remote Field Eddy Current Sensor. The sensor is installed on a mobile robot whose mission is the detection of anomalous regions in metal pipelines. Given the presence of noise that characterizes the data series, anomaly signals could be masked by noise and therefore difficult to identify in some instances. In order to enhance signal peaks that potentially identify anomalies we consider an entropy filter built on a-posteriori probability density functions associated with data series. Thresholds based on the Neyman-Pearson criterion for hypothesis testing are derived. The algorithmic tool is applied to the analysis of data from a portion of pipeline with a set of anomalies introduced at predetermined locations. Critical areas identifying anomalies capture the set of damaged locations, demonstrating the effectiveness of the filter in detection with Remote Field Eddy Current Sensor.

The number of solar panels deployed worldwide has rapidly increased. Solar panels are often placed in areas not easily accessible. It is also difficult for panel owners to be aware of their operating condition. Many environmental factors have negative effects on the efficiency of solar panels. To reduce the power lost caused by environmental factors, it is necessary to detect and classify the anomalous events occurring on the surface of solar panels. This thesis designs and studies a device to continuously measure the voltage output of solar panels and to transmit the time series data back to a personal computer using wireless communication. A program was developed to store and model this time series data. It also detected the existence of anomalies and classified the anomalies by modeling the data. In total, ten types of anomalies were considered. These anomaly types include temporary shading, permanent shading, fallen leaves, accumulating snow and melting snow among others. Previous time series anomaly detection algorithms do not perform well for reallife situations and are only capable of dealing with at most four different types of anomalies. In this work, a general mathematical model is proposed to give better performance in real-life test cases and to cover more than four types of anomalies. We note that the models can be generalized to detect and to classify anomalies for general time series data which is not necessarily generated from solar panel. We compared several techniques to detect and to classify anomalies including the auto-regressive integrated moving average model (ARIMA), neural networks, support vector machines and k-nearest-neighbors classification. We found that anomaly classification using the k-nearest-neighbors classification was able to accurately detect and classify 97% of the anomalies in our test set. The devices and algorithms have been tested with two small 12-volt solar panels.