No description available.

Thesis (MEng)--Stellenbosch University, 2015. / ENGLISH ABSTRACT: The Smart Grid is an initiative to make the existing utility grid more effective and efficient by making utility infrastructure smarter. The initiative affects all areas of the utility grid and all utility hardware. Communication to utility hardware for monitoring and remote configuration is central to the smart grid vision. The focus of this project is the Smart Transformer, a distribution transformer with built-in intelligence and communication capabilities. Data acquisition and remote configuration hardware and software was developed and installed on a distribution transformer for application in deep rural areas. The solution included communication capabilities and adheres to industry standards. The solution was tested and data acquisition and management were done using the OSIsoft PI System software. Field tests were performed to evaluate the effectiveness of the solution in a deep rural setting. It was found that the smart transformer can be effectively monitored, configured and controlled in a deep rural setting. The smart transformer concept was investigated in a microgrid context. The potential of a smart transformer within a microgrid was explored and the smart transformer as a microgrid market-enabler was focussed on. A simulation was performed to evaluate the role of a smart transformer as a microgrid market-enabling device. It was found that the smart transformer has the potential to serve as a market-enabling device. / AFRIKAANSE OPSOMMING: Die slim kragnetwerk is 'n initiatief om die bestaande kragnetwerk meer effektief en doeltreffend te maak deur kragnetwerk infrastruktuur se intelligensie te vermeerder. Die initiatief beïnvloed alle aspekte van die kragnetwerk en kragnetwerk hardeware. Kommunikasie met kragnetwerk hardeware vir moniteering en instelling oor 'n afstand is sentraal aan die slim kragnetwerk visie. Die fokus van hierdie projek is die slim transformator, 'n distribusie transformator met ingeboude intelligensie en kommunikasie vermoëns. Data verkryging en afstandelike instelling hardeware en sagteware was ontwikkel en installeer op 'n distribusie transformator vir toepasing in diep-landelike gebiede. Die oplossing sluit kommunikasie vermoëns in en voldoen aan industrie standaarde. Die oplossing was getoets en data verkryging en bestuur was geïmplementeer met gebruik van OSIsoft se PI Stelsel sagteware. Veldtoetse was gedoen om die effektiwiteit van die oplossing in diep-landelike gebiede te evalueer. Dit was gevind dat die slim transformator effektief gemoniteer, ingestel en beheer kan word in 'n diep-landelike omgewing. Die slim transformator konsep was ondersoek in 'n mikro-kragnetwerk konteks. Die potensiaal van 'n slim transformator binne 'n mikro-kragnetwerk was verken en die vermoë van 'n slim transformator om 'n mark binne 'n mikro-kragnetwerk in staat te stel was op gefokus.‘n Simulasie was uitgevoer om die vermoë wat 'n slim transformator het om 'n mark binne 'n mikro-kragnetwerk in staat te stel te evalueer. Dit was gevind dat 'n slim transformator die vermoë het om 'n mark binne 'n mirko-kragnetwerk in staat the stel.

The deep-sea hypersaline lakes in the Red Sea are among the most challenging, extreme, and unusual environments on the planet Earth. Despite their harshness to life, they are inhabited by diverse and novel members of prokaryotes. Methanogenesis was proposed as one of the main metabolic pathways that drive microbial colonization in similar habitats. However, not much is known about the identities of the methane-producing microbes in the Red Sea, let alone the way in which they could adapt to such poly extreme environments. Combining a range of microbial community assessment, cultivation and omics (genomics, transcriptomics, and single amplified genomics) approaches, this dissertation seeks to fill these gaps in our knowledge by studying archaeal composition, particularly methanogens, their genomic capacities and transcriptomic characteristics in order to elucidate their diversity, function, and adaptation to the deep-sea brines of the Red Sea. Although typical methanogens are not abundant in the samples collected from brine pool habitats of the Red Sea, the pilot cultivation experiment has revealed novel halophilic methanogenic species of the domain Archaea. Their physiological traits as well as their genomic and transcriptomic features unveil an interesting genetic and functional adaptive capacity that allows them to thrive in the unique deep-sea hypersaline environments in the Red Sea.

The aim of this doctoral thesis is to present a body of work aimed at reducing the time spent by animators manually constructing keyframed animation. To this end we present a number of state of the art machine learning techniques applied to the domain of character animation. Data-driven tools for the synthesis and production of character animation have a good track record of success. In particular, they have been adopted thoroughly in the games industry as they allow designers as well as animators to simply specify the high-level descriptions of the animations to be created, and the rest is produced automatically. Even so, these techniques have not been thoroughly adopted in the film industry in the production of keyframe based animation [Planet, 2012]. Due to this, the cost of producing high quality keyframed animation remains very high, and the time of professional animators is increasingly precious. We present our work in four main chapters. We first tackle the key problem in the adoption of data-driven tools for key framed animation - a problem called the inversion of the rig function. Secondly, we show the construction of a new tool for data-driven character animation called the motion manifold - a representation of motion constructed using deep learning that has a number of properties useful for animation research. Thirdly, we show how the motion manifold can be extended as a general tool for performing data-driven animation synthesis and editing. Finally, we show how these techniques developed for keyframed animation can also be adapted to advance the state of the art in the games industry.

abstract: Source selection is one of the foremost challenges for searching deep-web. For a user query, source selection involves selecting a subset of deep-web sources expected to provide relevant answers to the user query. Existing source selection models employ query-similarity based local measures for assessing source quality. These local measures are necessary but not sufficient as they are agnostic to source trustworthiness and result importance, which, given the autonomous and uncurated nature of deep-web, have become indispensible for searching deep-web. SourceRank provides a global measure for assessing source quality based on source trustworthiness and result importance. SourceRank's effectiveness has been evaluated in single-topic deep-web environments. The goal of the thesis is to extend sourcerank to a multi-topic deep-web environment. Topic-sensitive sourcerank is introduced as an effective way of extending sourcerank to a deep-web environment containing a set of representative topics. In topic-sensitive sourcerank, multiple sourcerank vectors are created, each biased towards a representative topic. At query time, using the topic of query keywords, a query-topic sensitive, composite sourcerank vector is computed as a linear combination of these pre-computed biased sourcerank vectors. Extensive experiments on more than a thousand sources in multiple domains show 18-85% improvements in result quality over Google Product Search and other existing methods. / Dissertation/Thesis / M.S. Computer Science 2011

abstract: This action research study is the culmination of several action cycles investigating cognitive information processing and learning strategies based on students approach to learning theory and assessing students' meta-cognitive learning, motivation, and reflective development suggestive of deep learning. The study introduces a reading assignment as an integrative teaching method with the purpose of challenging students' assumptions and requiring them to think from multiple perspectives thus influencing deep learning. The hypothesis is that students who are required to critically reflect on their own perceptions will develop the deep learning skills needed in the 21st century. Pre and post surveys were used to assess for changes in students' preferred approach to learning and reflective practice styles. Qualitative data was collected in the form of student stories and student literature circle transcripts to further describe student perceptions of the experience. Results indicate stories that include examples of critical reflection may influence students to use more transformational types of reflective learning actions. Approximately fifty percent of the students in the course increased their preference for deep learning by the end of the course. Further research is needed to determine the effect of narratives on student preferences for deep learning. / Dissertation/Thesis / Ed.D. Leadership and Innovation 2012

Pose estimation is the problem of detecting poses of people in images, multiperson pose estimation is the problem of detecting poses of multiple persons in images. This thesis investigates multi-person pose estimation by applying the associative embedding method on images from soccer videos. Three models are compared, first a pre-trained model, second a fine-tuned model and third a model extended to handle image sequences. The pre-trained method performed well on soccer images and the fine-tuned model performed better then the pre-trained model. The image sequence model performed equally as the fine-tuned model but not better. This thesis concludes that the associative embedding model is a feasible option for pose estimation in soccer videos and should be further researched.

The discovery of deep-sea chemosynthesis-based ecosystems including hydrothermal vents and cold seeps has greatly expanded our view of life on Earth. Nevertheless, for many benthic organisms in these ecosystems, little is known about where they come from, how scattered populations are connected by larval dispersal, and how they adapt to the local environments. Mussels of Bathymodiolus platifrons (Bivalvia: Mytilidae) are one of the dominant and foundation species in deep-sea chemosynthesis-based ecosystems. They are known to have a wide geographic distribution, and are also one of the few deep-sea species capable of living in both hydrothermal vents [in Okinawa Trough (OT)] and methane seeps [in the South China Sea (SCS) and Sagami Bay (SB)]. Previous population genetics studies of B. platifrons mostly relied on one to several genes, which suffered from the lack of sensitivity required to resolve their fine-scale genetic structure, and were unable to reveal their adaptation to the local environments. With the repaid development of molecular techniques, it is now possible to address their demographic mechanisms and local adaptation from a genome-wide perspective. Therefore, in the first part of my thesis, I aimed to generate genome-wide single nucleotide polymorphisms (SNPs) for B. platifrons via a combination of genome survey sequencing and the type IIB endonuclease restriction-site associated DNA (2b-RAD) approach, assess the potential use of SNPs in detecting fine-scale population genetic structure and signatures of diversifying selection, as well as their cross-species application in other bathymodioline mussels. Genome survey sequencing was conducted for one individual of B. platifrons. De novo assembly resulted in 781 720 sequences with a scaffold N50 of 2.9 kb. Using these sequences as a reference, 9307 genome-wide SNPs were identified from 28 B. platifrons individuals collected from a methane seep in the SCS and a hydrothermal vent in the middle OT (M-OT), with nine outlier SNPs showed significant evidence of diversifying selection. The small FST value (0.0126) estimated based on the neutral SNPs indicated high genetic connectivity between the two populations. However, the permutation test detected significant differences (P < 0.00001), indicating the two populations having clearly detectable genetic differentiation. The Bayesian clustering analyses and principle component analyses (PCA) performed based on either the neutral or outlier SNPs also showed that the two populations were genetically differentiated. This initial study successfully demonstrated the applicability of combining genome sequencing and 2b-RAD in population genomics studies of B. platifrons. Besides, using the survey genome of B. platifrons as a reference, a total of 10 199, 6429, and 3811 single nucleotide variants (SNVs) were detected from three bathymodioline mussels Bathymodiolus japonicus, Bathymodiolus aduloides, and Idas sp. These results highlighted the potential of cross-species and cross-genus applications of the B. platifrons genome for SNV/SNP identification among different bathymodioline lineages, which can be further used in various evolutionary and genetic studies. To have a deeper understanding of how individuals of B. platifrons are connected among and adapt to their habitats, in the second part of my thesis, I used both mitochondrial genes and genome-wide SNPs to conduct a more comprehensive population genetics/genomics study of B. platifrons. Three mitochondrial genes (i.e. atp6, cox1, and nad4) and 6398 SNPs generated by 2b-RAD were obtained from 110 B. platifrons individuals from six representative locations along their known distribution range in the Northwestern Pacific. The small FST values based on both types of genetic markers all revealed high genetic connectivity of B. platifrons, which may have been driven by the strong ocean currents (i.e. Kuroshio Current, North Pacific Intermediate Water). However, when using SNP datasets rather than mitochondrial genes, individuals in the SCS were identified as a distinct genetic group, indicating the Luzon Strait may serve as a dispersal barrier that limits their larval exchange between the SCS and the open area in the Northwestern Pacific. Moreover, a genetic subdivision of B. platifrons in the southern OT (S-OT) from those in M-OT and SB was observed when using 125 outlier SNPs for data analyses. The outlier-associated proteins were found to be involved in various biological processes, such as DNA and protein metabolism, transcription and translation, and response to stimulus, indicating local adaptation of B. platifrons even they are confronted with extensive gene flow in the OT-SB region. Furthermore, by using SNP datasets, populations in S-OT were revealed to be the source of gene flow to those in the SCS, M-OT, and SB. Overall, these results offered novel perspectives on the potential forces that may have led to the genetic differentiation and local adaptation of B. platifrons, which can serve as an example for other deep-sea species with high dispersal potential, and contribute to the designation of marine protected areas and conservation of deep-sea chemosynthesis-based ecosystems. Molluscan shell formation is one of the most common and abundant biomineralization processes in metazoans. Although composed of less than 5 wt% of the molluscan shells, shell matrix proteins (SMPs) are known to play multiple key roles during shell formation, such as providing a gel-like micro-environment to favour mineral precipitation, promoting crystal nucleation, as well as guiding and inhibiting crystal growth. To date, all studies on SMPs have focused on molluscs in terrestrial and shallow-water ecosystems with no reports for those living in the deep ocean. Herein, the third part of my thesis was to study the shell proteomes of B. platifrons and its shallow-water relative Modiolus philippinarum with the aim to bridge such knowledge gaps in biomineralization studies. A total of 94 and 55 SMPs were identified from the shell matrices of B. platifrons and M. philippinarum, respectively, with 31 SMPs shared between two species. These SMPs can be assigned into six broad categories, comprising calcium binding, polysaccharide interaction, enzyme, extracellular matrix-related proteins, immunity-related proteins, and those with uncharacterized functions. Many of them, such as tyrosinases, carbonic anhydrases, collagens, chitin-related proteins, peroxidases, as well as proteinase and proteinase inhibitor domain-containing proteins, have been widely found in molluscan shell matrices and other metazoan calcified tissues (e.g. exoskeletons of corals, tubes of tubeworms), whereas some others, such as cystatins, were found for the first time in molluscan shell matrices, and ferric-chelate reductase-like proteins and heme-binding proteins were to be detected for the first time in metazoan calcified tissues. This is the first report of the shell proteome of deep-sea molluscs, which will support various follow-up studies to better understand the functions of these SMPs, especially in relation to environmental adaptation. Overall, my population genetics/genomics studies have improved our understanding of the population dynamics, genetic connectivity, fine-scale genetic structure, and local adaptation of B. platifrons in the Northwestern Pacific, and my proteomics study has shed light on the biomineralization processes of molluscs in the deep ocean.

Increasing interest for classification of 3D geometrical data has led to discov- ery of PointNet, which is a neural network architecture capable of processing un- ordered point sets. This thesis explores several methods of utilizing conventional point features within PointNet and their impact on classification. Classification performance of the presented methods was experimentally evaluated and com- pared with a baseline PointNet model on four different datasets. The results of the experiments suggest that some of the considered features can improve clas- sification effectiveness of PointNet on difficult datasets with objects that are not aligned into canonical orientation. In particular, the well known spin image rep- resentations can be employed successfully and reliably within PointNet. Further- more, a feature-based alternative to spatial transformer, which is a sub-network of PointNet responsible for aligning misaligned objects into canonical orientation, have been introduced. Additional experiments demonstrate that the alternative might be competitive with spatial transformer on challenging datasets. 1

Depuis les années 1970, la télédétection a permis d’améliorer l’analyse de la surface de la Terre grâce aux images satellites produites sous format numérique. En comparaison avec les images aéroportées, les images satellites apportent plus d’information car elles ont une couverture spatiale plus importante et une période de revisite courte. L’essor de la télédétection a été accompagné de l’émergence des technologies de traitement qui ont permis aux utilisateurs de la communauté d’analyser les images satellites avec l’aide de chaînes de traitement de plus en plus automatiques. Depuis les années 1970, les différentes missions d’observation de la Terre ont permis d’accumuler une quantité d’information importante dans le temps. Ceci est dû notamment à l’amélioration du temps de revisite des satellites pour une même région, au raffinement de la résolution spatiale et à l’augmentation de la fauchée (couverture spatiale d’une acquisition). La télédétection, autrefois cantonnée à l’étude d’une seule image, s’est progressivement tournée et se tourne de plus en plus vers l’analyse de longues séries d’images multispectrales acquises à différentes dates. Le flux annuel d’images satellite est supposé atteindre plusieurs Péta octets prochainement. La disponibilité d’une si grande quantité de données représente un atout pour développer de chaines de traitement avancées. Les techniques d’apprentissage automatique beaucoup utilisées en télédétection se sont beaucoup améliorées. Les performances de robustesse des approches classiques d’apprentissage automatique étaient souvent limitées par la quantité de données disponibles. Des nouvelles techniques ont été développées pour utiliser efficacement ce nouveau flux important de données. Cependant, la quantité de données et la complexité des algorithmes mis en place nécessitent une grande puissance de calcul pour ces nouvelles chaînes de traitement. En parallèle, la puissance de calcul accessible pour le traitement d’images s’est aussi accrue. Les GPUs («Graphic Processing Unit ») sont de plus en plus utilisés et l’utilisation de cloud public ou privé est de plus en plus répandue. Désormais, pour le traitement d’images, toute la puissance nécessaire pour les chaînes de traitements automatiques est disponible à coût raisonnable. La conception des nouvelles chaînes de traitement doit prendre en compte ce nouveau facteur. En télédétection, l’augmentation du volume de données à exploiter est devenue une problématique due à la contrainte de la puissance de calcul nécessaire pour l’analyse. Les algorithmes de télédétection traditionnels ont été conçus pour des données pouvant être stockées en mémoire interne tout au long des traitements. Cette condition est de moins en moins respectée avec la quantité d’images et leur résolution. Les algorithmes de télédétection traditionnels nécessitent d’être revus et adaptés pour le traitement de données à grande échelle. Ce besoin n’est pas propre à la télédétection et se retrouve dans d’autres secteurs comme le web, la médecine, la reconnaissance vocale,… qui ont déjà résolu une partie de ces problèmes. Une partie des techniques et technologies développées par les autres domaines doivent encore être adaptées pour être appliquée aux images satellites. Cette thèse se focalise sur les algorithmes de télédétection pour le traitement de volumes de données massifs. En particulier, un premier algorithme existant d’apprentissage automatique est étudié et adapté pour une implantation distribuée. L’objectif de l’implantation est le passage à l’échelle c’est-à-dire que l’algorithme puisse traiter une grande quantité de données moyennant une puissance de calcul adapté. Enfin, la deuxième méthodologie proposée est basée sur des algorithmes récents d’apprentissage automatique les réseaux de neurones convolutionnels et propose une méthodologie pour les appliquer à nos cas d’utilisation sur des images satellites. / Since the 1970s, remote sensing has been a great tool to study the Earth in particular thanks to satellite images produced in digital format. Compared to airborne images, satellite images provide more information with a greater spatial coverage and a short revisit period. The rise of remote sensing was followed by the development of processing technologies enabling users to analyze satellite images with the help of automatic processing chains. Since the 1970s, the various Earth observation missions have gathered an important amount of information over time. This is caused in particular by the frequent revisiting time for the same region, the improvement of spatial resolution and the increase of the swath (spatial coverage of an acquisition). Remote sensing, which was once confined to the study of a single image, has gradually turned into the analysis of long time series of multispectral images acquired at different dates. The annual flow of satellite images is expected to reach several Petabytes in the near future. The availability of such a large amount of data is an asset to develop advanced processing chains. The machine learning techniques used in remote sensing have greatly improved. The robustness of traditional machine learning approaches was often limited by the amount of available data. New techniques have been developed to effectively use this new and important data flow. However, the amount of data and the complexity of the algorithms embedded in the new processing pipelines require a high computing power. In parallel, the computing power available for image processing has also increased. Graphic Processing Units (GPUs) are increasingly being used and the use of public or private clouds is becoming more widespread. Now, all the power required for image processing is available at a reasonable cost. The design of the new processing lines must take this new factor into account. In remote sensing, the volume of data currently available for exploitation has become a problem due to the constraint of the computing power required for the analysis. Traditional remote sensing algorithms have often been designed for data that can be stored in internal memory throughout processing. This condition is violated with the quantity of images and their resolution taken into account. Traditional remote sensing algorithms need to be reviewed and adapted for large-scale data processing. This need is not specific to remote sensing and is found in other sectors such as the web, medicine, speech recognition ... which have already solved some of these problems. Some of the techniques and technologies developed by the other domains still need to be adapted to be applied to satellite images. This thesis focuses on remote sensing algorithms for processing massive data volumes. In particular, a first algorithm of machine learning is studied and adapted for a distributed implementation. The aim of the implementation is the scalability, i.e. the algorithm can process a large quantity of data with a suitable computing power. Finally, the second proposed methodology is based on recent algorithms of learning convolutional neural networks and proposes a methodology to apply them to our cases of use on satellite images.