http://scholar.sun.ac.za:80/handle/10019.1/96336 Fri, 14 Jul 2017 14:01:56 GMT 2017-07-14T14:01:56Z http://scholar.sun.ac.za:80/handle/10019.1/100859 Concept-based exploration of rich semi-structured data collections Greene, Gillian J. ENGLISH ABSTRACT : Search has become one of the fundamental operations in computer science, allowing users to extract data and ultimately information from datasets. However, when users have no previous knowledge of a dataset, or have not clearly defined their search task and are therefore unable to formulate a direct query, their task becomes one of exploratory search or browsing rather than focused search or retrieval. While search and retrieval tools are widely provided, support for browsing of large and especially rich semi-structured datasets, is lacking. Semi-structured data is particularly difficult to explore and query because treating it as complete free-text causes a loss of important additional information which is encoded in the structured portions of the data while considering only the structured fields results in the loss of important free-text information. We therefore develop a framework to support exploration of semi-structured data, which is otherwise difficult to gather insights from, without requiring the user to have prior knowledge of the dataset or have formulated a specific query. Our approach uses a novel combination of tag clouds and concept lattices to facilitate data exploration, analysis and visualization by allowing the user to continuously update (i.e., add and remove) a set of keywords that the searched documents must contain. The document set is not directly provided as the result of a specific query, but aggregated information and properties of relevant documents are provided as a result. We apply our framework to data contained in software repositories, in two different ways for different goals to highlight the flexibility of the approach and the different tasks that can be supported using the same underlying dataset. We also instantiate our framework to support the exploration of a large collection of academic publication data. We evaluate the instantiations of our framework by conducting user and case studies, which indicate that our approach is usable and allows users to gather valuable information from semi-structured data archives.; AFRIKAANSE OPSOMMING : Soektogte is een van die fundamentele operasies in rekenaarwetenskap. Dit laat gebruikers toe om data, en uiteindelik inligting, vanuit datastelle te onttrek. Wanneer gebruikers egter geen vorige kennis van ’n datastel het nie, of hul soektog nie duidelik gedefinieer het nie, en dus nie in staat is om ’n direkte navraag te formuleer nie, word hul taak een van verkennende soek, of blaai, eerder as gefokusde soek of herwinning. Terwyl soeken herwinnings-instrumente algemeen beskikbaar is, ontbreek ondersteuning vir die verkenning van groot en veral ryk semi-gestruktureerde datastelle. Semi-gestruktureerde data is veral moeilik om te verken en na te vra omdat die hantering daarvan as slegs vrye teks ’n verlies van belangrike aanvullende inligting veroorsaak wat ingebou is in die gestruktureerde gedeeltes van die data, terwyl die inagneming van slegs die gestruktureerde velde weer lei tot ’n verlies van belangrike vrye teks inligting. Ons ontwikkel dus ’n raamwerk om die verkenning van semi-gestruktureerde data te ondersteun, wat andersins moeilik is om insigte uit te verkry, sonder om van die gebruiker te vereis dat hulle voorafgaande kennis van die datastel het, of ’n spesifieke navraag reeds geformuleer het. Ons benadering maak gebruik van ’n nuwe kombinasie van etiket-wolke en konsep-roosters om data-verkenning, data-analise, en data-visualisering te fasiliteer deur die gebruiker toe te laat om voortdurend ’n stel sleutelwoorde op te dateer (m.a.w. by te voeg of te verwyder) wat bevat moet word in die dokumente waarvoor gesoek word. Die dokument-stel word nie direk verskaf as die resultaat van ’n spesifieke navraag nie, maar saamgestelde inligting en eienskappe van relevante dokumente word eerder verskaf. Ons pas ons raamwerk toe op data wat in programmatuurargiewe gestoor is op twee verskillende maniere vir verskillende doelwitte om die buigsaamheid van die benadering en die verskillende take wat ondersteun kan word met behulp van dieselfde onderliggende datastel uit te lig. Ons instansieer ook ons raamwerk om die verkenning van ’n groot versameling van akademiese publikasiedata te ondersteun. Ons evalueer die instansies van ons raamwerk deur die gebruik van gebruiker en gevallestudies, wat daarop dui dat ons benadering bruikbaar is, en gebruikers in staat stel om waardevolle inligting vanuit semi-gestruktureerde data-argiewe in te samel. Thesis (PhD)--Stellenbosch University, 2017 Wed, 01 Mar 2017 00:00:00 GMT http://scholar.sun.ac.za:80/handle/10019.1/100859 2017-03-01T00:00:00Z http://scholar.sun.ac.za:80/handle/10019.1/100503 Unsupervised pre-training for fully convolutional neural networks Wiehman, Stiaan; Kroon, Steve; De Villiers, Hendrik Unsupervised pre-training of neural networks has been shown to act as a regularization technique, improving performance and reducing model variance. Recently, fully con-volutional networks (FCNs) have shown state-of-the-art results on various semantic segmentation tasks. Unfortunately, there is no efficient approach available for FCNs to benefit from unsupervised pre-training. Given the unique property of FCNs to output segmentation maps, we explore a novel variation of unsupervised pre-training specifically designed for FCNs. We extend an existing FCN, called U-net, to facilitate end-to-end unsupervised pre-training and apply it on the ISBI 2012 EM segmentation challenge data set. We performed a battery of significance tests for both equality of means and equality of variance, and show that our results are consistent with previous work on unsupervised pre-training obtained from much smaller networks. We conclude that end-to-end unsupervised pre-training for FCNs adds robustness to random initialization, thus reducing model variance. CITATION: Wiehman, S., Kroon, S. & De Villiers, H. 2016. Unsupervised pre-training for fully convolutional neural networks. Pattern Recognition Association of South Africa and Robotics and Mechatronics International Conference (PRASA-RobMech), 30 November-2 December 2016, Stellenbosch, South Africa.; The original publication is available at http://ieeexplore.ieee.org Fri, 01 Jan 2016 00:00:00 GMT http://scholar.sun.ac.za:80/handle/10019.1/100503 2016-01-01T00:00:00Z http://scholar.sun.ac.za:80/handle/10019.1/98844 Modelling the dynamics of the bitcoin blockchain Mwale, Mabvuto ENGLISH ABSTRACT : Bitcoin is a peer to peer (P2P) electronic payment system proposed by Nakamoto in 2008. Central to the operation of Bitcoin is the blockchain, which is, in essence, a public ledger of all transactions. The blockchain is maintained by a group of volunteers called miners, who are rewarded with bitcoins for successfully mining blocks. Mining a block involves collecting and validating transactions, adding validated transactions to a block, and finally adding the block to the blockchain and broadcasting the block to the peer network. The purpose of this thesis is to examine the profitability of block discarding attacks in the Bitcoin network. To this end, we developed a discrete event simulator to model the dynamics of the blockchain. We use a simplistic network model where all participants are miners. Initially, our investigation focusses on a model where all miners are honest, following the Bitcoin rules, and all block transmissions are subject to delays. We show that the delays cause forks to occur in the blockchain, which are quickly resolved. The simulation results for our network model closely agree with those observed from the real Bitcoin network. We then examine a block discarding attack referred to as selfish-mine [32], where it is claimed that a small group of colluding miners can subvert the Bitcoin protocol. We evaluate this claim using relative revenue (the fraction of the total revenue that is credited to the pool of colluding miners) and confirmed blocks (the absolute number of blocks credited to the pool of colluding miners at the end of a mining period) for the case where there is instantaneous block transmission, and the case where block transmission is subject to delays. We show that this claim is true for the first case. However, for the latter case, selfish-mine is only profitable (in terms of the relative revenue) when the pool commands more than 30% of the total computing power of the network. We further show that a higher relative revenue does not necessarily entail a larger number of confirmed blocks and that the presence of selfish-mine causes both the honest and the dishonest miners to fare worse. Finally we present a generalized form of selfish-mine, representing a greater variety of block discarding attacks, and use a genetic algorithm to search for an optimal configuration of the generalized selfish-mine that yields a better performance for the pool. We attempt to maximize either the relative revenue or the number of confirmed blocks. Our results show that the generalized form of selfish-mine when optimally configured yields better performance than standard selfish-mine, both in terms of the relative revenue and confirmed blocks.; AFRIKAANSE OPSOMMING : Bitcoin is 'n stelsel vir elektroniese betalings van portuur tot portuur (P2P) waarmee Nakamoto in 2008 vorendag gekom het. Die werking van Bitcoin draai om die blokketting, wat in wese 'n openbare grootboek van alle transaksies is. Die blokketting word bygehou deur 'n groep vrywilligers wat as myners bekend is, wat met bitcoins beloon word indien hulle blokke suksesvol ontgin. Die ontginning van 'n blok behels die insameling en stawing van transaksies, waarna gestaafde transaksies by 'n blok bygevoeg, die blok aan die blokketting gehaak en dan na die P2Pnetwerk versend word. Die doel van hierdie tesis is om die winsgewendheid van blokverwerpingsaanvalle in die Bitcoin-netwerk te ondersoek. Hiervoor is 'n diskrete gebeurtenissimuleerder ontwikkel om die dinamiek van die blokketting te modelleer. 'n Simplistiese netwerkmodel word gebruik waarin alle deelnemers myners is. Aanvanklik konsentreer die ondersoek op 'n model waarin alle myners eerlik is en die Bitcoin-reëls volg, en waarin alle blokversendings aan vertragings onderworpe is. Daar word aangetoon dat die vertragings vertakkings in die blokketting veroorsaak, wat vinnig uitgestryk word. Die simulasieresultate uit die netwerkmodel stem grootliks ooreen met wat in die werklike Bitcoinnetwerk waargeneem word. Daarna word 'n blokverwerpingsaanval wat as 'selfsugtige ontginning' bekend is [32], ondersoek, waar 'n klein groep myners na bewering kan saamspan om die Bitcoin-protokol te ondermyn. Hierdie bewering word beoordeel aan die hand van relatiewe inkomste (die fraksie van totale inkomste wat aan die betrokke groep myners toegeskryf kan word) en bevestigde blokke (die absolute getal blokke wat aan die einde van 'n ontginningstydperk aan die groep myners toegeskryf kan word) in 'n scenario van onmiddellike blokversending sowel as 'n scenario waar blokversending aan vertragings onderworpe is. Daar word aangetoon dat hierdie bewering waar is in die eerste scenario. In die tweede scenario is selfsugtige ontginning slegs winsgewend (wat relatiewe inkomste betref) indien die groep meer as 30% van die algehele berekeningsvermoë van die netwerk beheer. Die studie toon ook dat 'n hoër relatiewe inkomste nie noodwendig 'n groter getal bevestigde blokke behels nie, en dat die aanwesigheid van selfsugtige ontginning eerlike sowel as oneerlike myners se prestasie verswak. Laastens word 'n veralgemeende vorm van selfsugtige ontginning aangebied wat 'n groter verskeidenheid blokverwerpingsaanvalle voorstel. 'n Genetiese algoritme word gebruik om 'n optimale konfigurasie van die veralgemeende selfsugtige ontginning te bepaal wat 'n beter prestasie vir die groep oplewer. Hiervoor probeer die navorser hetsy die relatiewe inkomste of die getal bevestigde blokke maksimaliseer. Die resultate toon dat die veralgemeende vorm van selfsugtige ontginning in die optimale konfigurasie beter prestasie as standaardselfsugtige ontginning oplewer wat relatiewe inkomste sowel as bevestigde blokke betref. Thesis (MSc)--Stellenbosch University, 2016 Tue, 01 Mar 2016 00:00:00 GMT http://scholar.sun.ac.za:80/handle/10019.1/98844 2016-03-01T00:00:00Z http://scholar.sun.ac.za:80/handle/10019.1/98838 Detecting and quantifying resource contention in concurrent programs Venter, Dirk Willem ENGLISH ABSTRACT : Parallel programs, both shared-memory and message-passing programs, typically require the sharing of resources. For example, software resources, such as shared mutual exclusion locks and hardware resources, such as caches and memory. Shared resources can only be used by one thread or process at a time. The competition for limited resources is called resource contention. The result of resource contention is delays while waiting for access to a resource and/or extra computational overhead to resolve the request for a resource. Thus, the performance of the program can be improved by identifying and reducing contention for shared resources. This study investigates the effect of individual types of contention for hardware and software resources in detail and discusses the three tools that were developed to identify and quantify the sources of contention in concurrent programs.; AFRIKAANSE OPSOMMING : Geen Afrikaanse opsomming geskikbaar nie Thesis (MSc)--Stellenbosch University, 2016 Tue, 01 Mar 2016 00:00:00 GMT http://scholar.sun.ac.za:80/handle/10019.1/98838 2016-03-01T00:00:00Z http://scholar.sun.ac.za:80/handle/10019.1/98818 Automatic Prediction of Comment Quality Brand, Dirk Johannes ENGLISH ABSTRACT : The problem of identifying and assessing the quality of short texts (e.g. comments, reviews or web searches) has been intensively studied. There are great bene ts to being able to analyse short texts. As an example, advertisers might be interested in the sentiment of product reviews on e-commerce sites to more e ciently pair marketing material to content. Analysing short texts is a di cult problem, because traditional machine learning models generally perform better on data sets with larger samples, which often translates to more features. More data allow for better estimation of parameters for these models. Short texts generally do not have much content, but still carry high variability in that they may still consist of a large corpus of words. This thesis investigates various methods for feature extraction for short texts in the context of online user comments. These methods include the leading manual feature extraction techniques for short texts, N-gram models and techniques based on word embeddings. The e ect of using di erent kernels for a support vector classi er is also investigated. The investigation is centred around two data sets, one provided by News24 and the other extracted from Slashdot.org. It was found that N-gram models performed relatively well, mostly outperforming manual feature extraction techniques.; AFRIKAANSE OPSOMMING : Om die kwaliteit van kort tekste (bv. internet kommentaar, soektogte of resensies) te identi seer en te analiseer, is 'n probleem wat al redelik sorgvuldig in die navorsing bestudeer is. Daar is baie te baat by die vermo ë om die kwaliteit van aanlyn teks te analiseer. Byvoorbeeld, aanlyn winkels mag moontlik geinteresseerd wees in die sentiment van die verbruikers wat produkresensies gee oor hul produkte, aangesien dit kan help om meer akkurate bemarkings materiaal vir produkte te genereer. Analise van kort tekste is 'n uitdagende probleem, want tradisionele masjienleer algoritmes vaar gewoonlik beter op datastelle met meer kernmerke as wat kort tekste kan bied. Ryker datastelle laat toe vir meer akkurate skatting van model parameters. Hierdie tesis bestudeer verskeie metodes vir kenmerkkonstruksie van kort tekste in die konteks van aanlyn kommentaar. Die metodes sluit die voorstaande handgemaakde kenmerkkonstruksie tegnieke vir kort tekste, N-gram modelle en woordinbeddinge in. Die e ek van verskillende kernmetodes vir klassi kasie modelle word ook bestudeer. Die studie is gefokus rondom twee datastelle waarvan een deur News24 voorsien is en die ander vanaf Slashdot. org bekom is. Ons het gevind that N-gram modelle meestal beter presteer as die handgemaakde kenmerkkonstruksie tegnieke. Thesis (MSc)--Stellenbosch University, 2016 Tue, 01 Mar 2016 00:00:00 GMT http://scholar.sun.ac.za:80/handle/10019.1/98818 2016-03-01T00:00:00Z http://scholar.sun.ac.za:80/handle/10019.1/97814 Automatic recognition and interpretation of finite state automata diagrams Babalola, Olusola Tope ENGLISH ABSTRACT : An application capable of reading graphically-encoded information is beneficial to blind or visually impaired students. Such a system needs to recognize and understand visual markings and their arrangement as presented in a diagram image. In that light, this thesis examines the practical possibility of a real world system for the automatic recognition and interpretation of machine-printed Finite State Automata diagrams. The suggested system uses known image processing and pattern recognition methods to extract the visual markings from the diagram image pixels. A second stage, to interpret the meaning of the diagram, is based on modeling the language of Finite State Automata diagrams using Constraint Multiset Grammars. Our results show that a practical application for automatic interpretation of Finite State Automata diagrams is possible.; AFRIKAANSE OPSOMMING : Geen Afrikaanse opsomming geskikbaar nie Thesis (MSc)--Stellenbosch University, 2015 Tue, 01 Dec 2015 00:00:00 GMT http://scholar.sun.ac.za:80/handle/10019.1/97814 2015-12-01T00:00:00Z http://scholar.sun.ac.za:80/handle/10019.1/97081 Impendulo: A Tool for Analysing Programmer Behaviour Jordaan, Pieter ENGLISH ABSTRACT : Automated submission systems for Computer Science courses are common- place today, however, these systems are typically solely focused on grading submissions and their ability to provide analysis and feedback is vastly under- utilised. This thesis seeks to address this by presenting Impendulo, a platform for analysing programmer behaviour. Impendulo allows one to study student per- formance at both a group and an individual level. This is achieved through the use of a customisable set of software analysis tools which Impendulo runs on user submissions. The results produced by the tools are parsed by Impendulo to produce rich feedback in the form of interactive charts, annotated code and detailed reports. In order to ascertain whether Impendulo's approach is viable, experimental trials were conducted with it throughout its development process. These trials consisted of a group of students using the Impendulo system while solving a set of programming problems. After each trial, the results were studied and used to determine which aspects of the system needed to be improved. At the end of Impendulo's development, all of the experiments were studied again to gain insight into the tested students' programming.; AFRIKAANSE OPSOMMING : Outomatiese inhandigings sisteme vir rekenaarwetenskap kursusse is vandag 'n algemene verskynsel, maar hierde sisteme is tipies net gefokus daarop om werk te gradeer en hulle kapasiteit om analise en terugvoer te lewer word nie benut nie. Hierdie tesis beoog om hierdie kwessie aan te pak met die ontwikkeling van Impendulo, 'n platform vir die analise van programmeerder gedrag. Impendulo laat gebruikers toe om studente se vordering te monitor op beide 'n individele en 'n groep vlak. Dit word bereik deur 'n student se werk te analiseer met 'n aanpasbare stel analise sagteware. Die resultate wat deur die sagteware geproduseer word, word deur Impendulo ontleed sodat dit terugvoer kan gee in die vorm van interaktiewe gra eke, geannoteerde kode en gedetailleerde verslagte. Eksperimente is geloop tydens Impendulo se ontwikkeling om te bepaal of ons benadering prakties is. Hierdie eksperimente het bestaan uit 'n groep van studente wat die Impendulo sisteem gebruik terwyl hulle aan 'n stel probleme werk. Na elke eksperiment is die resultate daarvan ontleed en gebruik om te bepaal watter aspekte van die sisteem aandag benodig. Aan die einde van Impendulo se ontwikkeling is al die eksperimente weer na gegaan om insig te kry oor hoe die getoetste studente programmeer. Thesis (MSc)--Stellenbosch University, 2015. Sun, 01 Mar 2015 00:00:00 GMT http://scholar.sun.ac.za:80/handle/10019.1/97081 2015-03-01T00:00:00Z http://scholar.sun.ac.za:80/handle/10019.1/96961 A parallel cellular automaton simulation framework using CUDA Fourie, Ryno ENGLISH ABSTRACT : In the current digital age, the use of cellular automata to simulate natural systems has grown more popular as our understanding of cellular systems increases. Up until about a decade ago, digital models based on the concept of cellular automata have primarily been simulated with sequential rule application algorithms, which do not exploit the inherent parallel nature of cellular automata. However, since parallel computation platforms have become more commercially available, researchers have started to investigate the advantages of parallel rule application algorithms for basic cellular automata. For this thesis, a parallel cellular automaton framework, based on NVIDIA CUDA is developed to simplify the implementation of a wide range of cellular automata. This framework is used to investigate the potential performance advantages of using graphical processing units as a parallel processing platform for cellular automata.; AFRIKAANSE OPSOMMING : In die huidige digitale era het die gebruik van sellulêre outomate om natuurlike stelsels te simuleer, aansienlik toegeneem soos wat ons begrip van sellulêre stelsels verbreed word. Tot om en by 'n dekade gelede is digitale modelle wat met behulp van sellulêre outomate gesimuleer word, hoofsaaklik met sekwensiële reëlfunksies gesimuleer. As gevolg hiervan het die inherente parallelle natuur van sellulêre outomate nie tot sy volle reg gekom nie. Aangesien parallelle berekenings-platforms egter onlangs meer kommersieël beskikbaar geraak het, span navorsers hierdie platforms nou in om parallelle reëlfunksies te skep vir meer basiese sellulêre outomate. Vir hierdie tesis is 'n parallelle sellulêre outomaat simulasieraamwerk geskep, wat gebruik maak van die NVIDIA CUDA parallelle berekenings-platform. Hierdie raamwerk is geskep om die implementasie van 'n verskeidenheid van sellulêre outomate te vereenvoudig, en is ingespan om die potensiële tydsvoordeel van gra ese verwerkingseenhede te ondersoek in die implementasie van sellulêre outomate. Thesis (MSc)--Stellenbosch University, 2015. Sun, 01 Mar 2015 00:00:00 GMT http://scholar.sun.ac.za:80/handle/10019.1/96961 2015-03-01T00:00:00Z http://scholar.sun.ac.za:80/handle/10019.1/96853 Combining reverse debugging and live programming towards visual thinking in computer programming Coetzee, Abraham Liebrecht ENGLISH ABSTRACT : Interaction plays a key role in the process of learning, and a learner’s abilities are enhanced when multiple cognitive functions work in parallel, especially those related to language and visuals. Time is the most fundamental variable that governs the interaction between programmer and computer, and the substantial temporal separation of cause and effect leads to poor mental models. Furthermore, programmers do not have means by which to express their mental models. The feasibility of combining reverse debugging and live programming was therefore investigated. This combination was found to be feasible, and a reverse debugger with higher levels of liveness was created for the Python programming language. It establishes a foundation for combining language and visual models as aids in computer programming education.; AFRIKAANSE OPSOMMING : Interaksie speel ’n belangrike rol in die proses van leer, en ’n leerder se vermoëns verbeter wanneer verskeie kognitiewe funksies in parallel opereer, veral dié wat verwant is aan taal en visuele denke. Tyd is die mees fundamentele veranderlike wat die interaksie tussen programmeerder en rekenaar reguleer, en die aansienlike temporele skeiding tussen oorsaak en gevolg lei tot swak kognitiewe modelle. Programmeerders het boonop nie middelle om kognitiewe modelle te artikuleer nie. Die uitvoerbaarheid van ’n kombinasie van terug-in-tyd ontfouting en lewendige programmering was daarom ondersoek. Daar was bevind dat so ’n kombinasie moontlik is, en ’n terug-in-tyd ontfouter met hoër vlakke van lewendigheid was geskep vir die Python programmeringstaal. Dit vestig ’n fondament om taal en visuele modelle te kombineer as hulpmiddels in rekenaarprogrammering onderwys. Thesis (MSc)--Stellenbosch University, 2015 Sun, 01 Mar 2015 00:00:00 GMT http://scholar.sun.ac.za:80/handle/10019.1/96853 2015-03-01T00:00:00Z