Dissertations / Theses on the topic 'Acceleration patterns'

The aim of this work was to develop devices for elite athletes to record performance related parameters during their training. A device was initially designed and built for rowing to record the motion of the boat. This was to gain understanding of motion signals in a one dimensional plane. The device uses a iPAQ handheld computer for recording and display of data to the user. Using the knowledge obtained from the accelerometer data of the rowing system an initial prototype device was designed and constructed for use in swimming. This device was required to be wearable whilst the swimmer was training, thus it had to record the data onboard. A second version of the swimming device was constructed to improve the usability of the device. The swimming device has fully sealed electronics, wireless charging and infrared communications. The device records three dimensional acceleration patterns at 150Hz, and can store over 6 hours of data using the internal memory. The device can operate for greater than 12 hours before needing to be recharged. The data collected from the swimming device was used to develop processing algorithms to extract when the swimmers push off from the wall, the type of stroke they are swimming, and for freestyle the stroke count. The results of the wall push off algorithm were compared against manual hand timing with 90% algorithm results being with ±1 second of the hand timing data. The stroke type identification algorithm determines which stroke is being swum and presently has an accuracy of 95%. The results of the freestyle stroke count algorithm were compared against manual stroke counts from raw accelerometers data and underwater video. Of the 164 data sets analysed over 90% of the algorithm results were within ±1 strokes of the manual recorded stroke counts.

The aim of this work was to develop devices for elite athletes to record performance related parameters during their training. A device was initially designed and built for rowing to record the motion of the boat. This was to gain understanding of motion signals in a one dimensional plane. The device uses a iPAQ handheld computer for recording and display of data to the user. Using the knowledge obtained from the accelerometer data of the rowing system an initial prototype device was designed and constructed for use in swimming. This device was required to be wearable whilst the swimmer was training, thus it had to record the data onboard. A second version of the swimming device was constructed to improve the usability of the device. The swimming device has fully sealed electronics, wireless charging and infrared communications. The device records three dimensional acceleration patterns at 150Hz, and can store over 6 hours of data using the internal memory. The device can operate for greater than 12 hours before needing to be recharged. The data collected from the swimming device was used to develop processing algorithms to extract when the swimmers push off from the wall, the type of stroke they are swimming, and for freestyle the stroke count. The results of the wall push off algorithm were compared against manual hand timing with 90% algorithm results being with ±1 second of the hand timing data. The stroke type identification algorithm determines which stroke is being swum and presently has an accuracy of 95%. The results of the freestyle stroke count algorithm were compared against manual stroke counts from raw accelerometers data and underwater video. Of the 164 data sets analysed over 90% of the algorithm results were within ±1 strokes of the manual recorded stroke counts.
Thesis (Masters)
Master of Philosophy (MPhil)
School of Microelectronic Engineering
Full Text

L'objectiu d'aquesta tesis es crear o adaptar models de programació per a fer els processadors multi-core accessibles per a la majoria de programadors. Aquest objectiu inclou la possibilitat de reusar els algoritmes existents, la capacitat de depuració, I la capacitat d'introduir els canvis de forma incremental. Contrastem les solucions proposades en diversos tipus de multi-core, incloent sistemes homogenis i heterogenis, i sistemes de memòria compartida i memòria distribuïda. A més a més contribuïm exposant algorismes i programes reals i ensenyant com aquests es poden ser usat per aplicacions en temps quasi real.
The aim of this thesis is to create or adapt a programming model in order to make multi-core processors accessible by almost every programmer. This objective includes existing codes and algorithms reuse, debuggability, and the capacity to introduce changes incrementally. We face multi-cores with many architectures including homogeneity versus heterogeneity and shared-memory versus distributed-memory. We also contribute by exposing real algorithms and programs and showing how some of them can be used for quasi realtime applications.

Background: String searching within a large corpus of data is a critical component of digital forensic (DF) analysis techniques such as file carving. The continuing increase in capacity of consumer storage devices requires similar improvements to the performance of string searching techniques employed by DF tools used to analyse forensic data. As string searching is a trivially-parallelisable problem, general purpose graphic processing unit (GPGPU) approaches are a natural fit. Currently, only some of the research in employing GPGPU programming has been transferred to the field of DF, of which, a closed-source GPGPU framework was used— Complete Unified Device Architecture (CUDA). Findings from these earlier studies have found that local storage devices from which forensic data are read present an insurmountable performance bottleneck. Aim: This research hypothesises that modern storage devices no longer present a performance bottleneck to the currently used processing techniques of the field, and proposes that an open-standards GPGPU framework solution – Open Computing Language (OpenCL) – would be better suited to accelerate file carving with wider compatibility across an array of modern GPGPU hardware. This research further hypothesises that a modern multi-string searching algorithm may be better adapted to fulfil the requirements of DF investigation. Methods: This research presents a review of existing research and tools used to perform file carving and acknowledges related work within the field. To test the hypothesis, parallel file carving software was created using C# and OpenCL, employing both a traditional string searching algorithm and a modern multi-string searching algorithm to conduct an analysis of forensic data. A set of case studies that demonstrate and evaluate potential benefits of adopting various methods in conducting string searching on forensic data are given. This research concludes with a final case study which evaluates the performance to perform file carving with the best-proposed string searching solution and compares the result with an existing file carving tool— Foremost. Results: The results demonstrated from the research establish that utilising the parallelised OpenCL and Parallel Failureless Aho-Corasick (PFAC) algorithm solution demonstrates significantly greater processing improvements from the use of a single, and multiple, GPUs on modern hardware. In comparison to CPU approaches, GPGPU processing models were observed to minimised the amount of time required to search for greater amounts of patterns. Results also showed that employing PFAC also delivers significant performance increases over the BM algorithm. The method employed to read data from storage devices was also seen to have a significant effect on the time required to perform string searching and file carving. Conclusions: Empirical testing shows that the proposed string searching method is believed to be more efficient than the widely-adopted Boyer-Moore algorithms when applied to string searching and performing file carving. The developed OpenCL GPGPU processing framework was found to be more efficient than CPU counterparts when searching for greater amounts of patterns within data. This research also refutes claims that file carving is solely limited by the performance of the storage device, and presents compelling evidence that performance is bound by the combination of the performance of the storage device and processing technique employed.

Die folgende Arbeit befasst sich mit der Alltagstauglichkeit moderner Massendatenverarbeitung, insbesondere mit dem Problem der Assoziationsregelanalyse. Vorhandene Datenmengen wachsen stark an, aber deren Auswertung ist für ungeübte Anwender schwierig. Daher verzichten Unternehmen auf Informationen, welche prinzipiell vorhanden sind. Assoziationsregeln zeigen in diesen Daten Abhängigkeiten zwischen den Elementen eines Datenbestandes, beispielsweise zwischen verkauften Produkten. Diese Regeln können mit Interessantheitsmaßen versehen werden, welche dem Anwender das Erkennen wichtiger Zusammenhänge ermöglichen. Es werden Ansätze gezeigt, dem Nutzer die Auswertung der Daten zu erleichtern. Das betrifft sowohl die robuste Arbeitsweise der Verfahren als auch die einfache Auswertung der Regeln. Die vorgestellten Algorithmen passen sich dabei an die zu verarbeitenden Daten an, was sie von anderen Verfahren unterscheidet. Assoziationsregelsuchen benötigen die Extraktion häufiger Kombinationen (EHK). Hierfür werden Möglichkeiten gezeigt, Lösungsansätze auf die Eigenschaften moderne System anzupassen. Als Ansatz werden Verfahren zur Berechnung der häufigsten $N$ Kombinationen erläutert, welche anders als bekannte Ansätze leicht konfigurierbar sind. Moderne Systeme rechnen zudem oft verteilt. Diese Rechnerverbünde können große Datenmengen parallel verarbeiten, benötigen jedoch die Vereinigung lokaler Ergebnisse. Für verteilte Top-N-EHK auf realistischen Partitionierungen werden hierfür Ansätze mit verschiedenen Eigenschaften präsentiert. Aus den häufigen Kombinationen werden Assoziationsregeln gebildet, deren Aufbereitung ebenfalls einfach durchführbar sein soll. In der Literatur wurden viele Maße vorgestellt. Je nach den Anforderungen entsprechen sie je einer subjektiven Bewertung, allerdings nicht zwingend der des Anwenders. Hierfür wird untersucht, wie mehrere Interessantheitsmaßen zu einem globalen Maß vereinigt werden können. Dies findet Regeln, welche mehrfach wichtig erschienen. Der Nutzer kann mit den Vorschlägen sein Suchziel eingrenzen. Ein zweiter Ansatz gruppiert Regeln. Dies erfolgt über die Häufigkeiten der Regelelemente, welche die Grundlage von Interessantheitsmaßen bilden. Die Regeln einer solchen Gruppe sind daher bezüglich vieler Interessantheitsmaßen ähnlich und können gemeinsam ausgewertet werden. Dies reduziert den manuellen Aufwand des Nutzers. Diese Arbeit zeigt Möglichkeiten, Assoziationsregelsuchen auf einen breiten Benutzerkreis zu erweitern und neue Anwender zu erreichen. Die Assoziationsregelsuche wird dabei derart vereinfacht, dass sie statt als Spezialanwendung als leicht nutzbares Werkzeug zur Datenanalyse verwendet werden kann
The importance of data mining is widely acknowledged today. Mining for association rules and frequent patterns is a central activity in data mining. Three main strategies are available for such mining: APRIORI , FP-tree-based approaches like FP-GROWTH, and algorithms based on vertical data structures and depth-first mining strategies like ECLAT and CHARM. Unfortunately, most of these algorithms are only moderately suitable for many “real-world” scenarios because their usability and the special characteristics of the data are two aspects of practical association rule mining that require further work. All mining strategies for frequent patterns use a parameter called minimum support to define a minimum occurrence frequency for searched patterns. This parameter cuts down the number of patterns searched to improve the relevance of the results. In complex business scenarios, it can be difficult and expensive to define a suitable value for the minimum support because it depends strongly on the particular datasets. Users are often unable to set this parameter for unknown datasets, and unsuitable minimum-support values can extract millions of frequent patterns and generate enormous runtimes. For this reason, it is not feasible to permit ad-hoc data mining by unskilled users. Such users do not have the knowledge and time to define suitable parameters by trial-and-error procedures. Discussions with users of SAP software have revealed great interest in the results of association-rule mining techniques, but most of these users are unable or unwilling to set very technical parameters. Given such user constraints, several studies have addressed the problem of replacing the minimum-support parameter with more intuitive top-n strategies. We have developed an adaptive mining algorithm to give untrained SAP users a tool to analyze their data easily without the need for elaborate data preparation and parameter determination. Previously implemented approaches of distributed frequent-pattern mining were expensive and time-consuming tasks for specialists. In contrast, we propose a method to accelerate and simplify the mining process by using top-n strategies and relaxing some requirements on the results, such as completeness. Unlike such data approximation techniques as sampling, our algorithm always returns exact frequency counts. The only drawback is that the result set may fail to include some of the patterns up to a specific frequency threshold. Another aspect of real-world datasets is the fact that they are often partitioned for shared-nothing architectures, following business-specific parameters like location, fiscal year, or branch office. Users may also want to conduct mining operations spanning data from different partners, even if the local data from the respective partners cannot be integrated at a single location for data security reasons or due to their large volume. Almost every data mining solution is constrained by the need to hide complexity. As far as possible, the solution should offer a simple user interface that hides technical aspects like data distribution and data preparation. Given that BW Accelerator users have such simplicity and distribution requirements, we have developed an adaptive mining algorithm to give unskilled users a tool to analyze their data easily, without the need for complex data preparation or consolidation. For example, Business Intelligence scenarios often partition large data volumes by fiscal year to enable efficient optimizations for the data used in actual workloads. For most mining queries, more than one data partition is of interest, and therefore, distribution handling that leaves the data unaffected is necessary. The algorithms presented in this paper have been developed to work with data stored in SAP BW. A salient feature of SAP BW Accelerator is that it is implemented as a distributed landscape that sits on top of a large number of shared-nothing blade servers. Its main task is to execute OLAP queries that require fast aggregation of many millions of rows of data. Therefore, the distribution of data over the dedicated storage is optimized for such workloads. Data mining scenarios use the same data from storage, but reporting takes precedence over data mining, and hence, the data cannot be redistributed without massive costs. Distribution by special data semantics or user-defined selections can produce many partitions and very different partition sizes. The handling of such real-world distributions for frequent-pattern mining is an important task, but it conflicts with the requirement of balanced partition

The main purpose of this project is to investigate different tool steels in terms of their ability to withstand material transfer buildup, so-called galling, occurring in SMF (sheet metal forming) operations. The ability to withstand galling is vital to optimize cost-effectiveness and increase the work tool’s effective operational time. This investigation studies four different tool steels, including a TiN-coating, with the intention of evaluating the microstructures, chemical composition and hardness effect on galling resistance in dry conditions using a slider-on-flatsurface (SOFS) tribo-tester which measures the coefficient of friction during sliding.

An OP (optical profilometer) was used to measure the size and geometry of lump growth on the tool and damage on the work sheet. A scanning electron microscope (SEM) was used to identify the interacting tribological mechanisms exhibited at different stages during the slide. The SEM figures confirmed three different types of characteristic patterns exhibited in the tracks after tribo- testing which were categorized as mild adhesive, abrasive and severe adhesive damage.

A SEM figure that illustrates a ragged contact surface and an obvious change in the sheet materials plastic behavior is in this report regarded as a sign of severe adhesive contact, the characteristics could possibly be explained by local high temperature and high pressure followed by a sudden pressure drop and creation of hardened welds or solders between the two surfaces which increase the frictional input needed for further advancement. Friction coefficients observed in the initial 100% mild adhesive stage were, μ=0,22-0,26 succeeded by abrasive SEM characteristics often in association with mild adhesive contact and friction values between μ=0,25-0,4 which where sometimes followed by severe adhesive SEM characteristics in 100% of the contact zone with friction values between μ=0,34- 0,9 respectively. The tool material that performed best according to the friction detection criteria was Sv21 closely followed by Sleipner (TiN coated) and Va40 (HRC 63.3). Unfortunately was the friction criteria, a significant raise in friction for defining a sliding length to galling, not adequate for dry conditions due to immediate material transfer succeeded by cyclic changes between partial or 100% abrasive+mild adhesive and severe adhesive contact. The mechanism that change abrasive wear in association with mild adhesive contact, (moderate friction input), to sever adhesive wear, (higher friction input), is dependent on lump shape (lump geometry) and can appear at comparably low speeds 0,04-0,08 [m/s] and low friction energy input (μ=0,34), the magnitude of the change in friction is therefore not always significant and hardly detectable on the friction graph. This was quite unexpected but could be explained by concentration of friction energy rater than the absolute amount. The problem with using friction graphs for galling evaluation was increased even further when a very small lump size and low corresponding rate of material transfer to the tool surface caused a sustainable high raise in friction (μ≈0,3→0,6) on a TiN-coated tool steel called Sleipner.

A hardly detectable or similar friction raise for Sv21 and Va40 showed much larger corresponding lump size and rate of material transfer. This means that friction graphs demonstrate a clear problem with quantifying lump size [m3] and rate of  material transfer [m3/s]. Another phenomenon called stick slip behavior, material transfer and lump growth followed by a sudden decrease in lump size and transfer of material back to the work sheet, is also not possible to detect on a friction graph. Because a drop in friction can easily be a change in contact temperature and lump attack angle due to a growing lump and not a decreasing lump.

 

The conclusion, a friction graph is not suited for galling evaluation and ranking in dry SOFS conditions. A ranking should primarily be based on dimensional OP measurements of the cross section of formed tracks and scratches or preferably by repeated OP measurements of the tool surface during a single test, the last revel the exact lump growth history and true lump growth even in the sliding direction.

 

civilingenjörsexamen

Steel-deck composite floor systems are being increasingly used in high-rise building construction, especially in Australia, as they are economical and easy to construct. These composite floor systems use high strength materials to achieve longer spans and are thus slender. As a result, they are vulnerable to vibration induced under service loads. These floors are normally designed using static methods which will not reveal the true behaviour and miss the dynamic amplifications resulting in inappropriate designs, which ultimately cause vibration and discomfort to occupants. At present there is no adequate design guidance to address the vibration in these composite floors, due to a lack of research information, resulting in wasteful post event retrofits. To address this gap in knowledge, a comprehensive research project is presented in this thesis, which investigated the dynamic performance of composite floors under various human induced loads. A popular type of composite floor system was selected for this investigation and subjected to load models representing different human activities. These load models have variable parameters such as load intensity, activity type (contact ratio), activity frequency and damping and are applied as pattern loads to capture the maximum responses in terms of deflections and accelerations. Computer models calibrated against experimental results are used in the analysis to generate the required information. The dynamic responses of deflections and accelerations are compared with the serviceability deflection limits and human comfort levels (of accelerations) to assess these floor types. This thesis also treats the use of visco-elastic (VE) dampers to mitigate excessive vibrations in steel-deck composite floors. VE damper properties have been presented and their performances in reducing the excessive vibrations have been assessed this thesis. The results identified possible occupancies under different loading conditions that can be used in planning, design and evaluation. The findings can also be used to plan retrofitting measures in problematic floor systems.

Steel-deck composite floor systems are being increasingly used in high-rise building construction, especially in Australia, as they are economical and easy to construct. These composite floor systems use high strength materials to achieve longer spans and are thus slender. As a result, they are vulnerable to vibration induced under service loads. These floors are normally designed using static methods which will not reveal the true behaviour and miss the dynamic amplifications resulting in inappropriate designs, which ultimately cause vibration and discomfort to occupants. At present there is no adequate design guidance to address the vibration in these composite floors, due to a lack of research information, resulting in wasteful post event retrofits. To address this gap in knowledge, a comprehensive research project is presented in this thesis, which investigated the dynamic performance of composite floors under various human induced loads. A popular type of composite floor system was selected for this investigation and subjected to load models representing different human activities. These load models have variable parameters such as load intensity, activity type (contact ratio), activity frequency and damping and are applied as pattern loads to capture the maximum responses in terms of deflections and accelerations. Computer models calibrated against experimental results are used in the analysis to generate the required information. The dynamic responses of deflections and accelerations are compared with the serviceability deflection limits and human comfort levels (of accelerations) to assess these floor types. This thesis also treats the use of visco-elastic (VE) dampers to mitigate excessive vibrations in steel-deck composite floors. VE damper properties have been presented and their performances in reducing the excessive vibrations have been assessed this thesis. The results identified possible occupancies under different loading conditions that can be used in planning, design and evaluation. The findings can also be used to plan retrofitting measures in problematic floor systems.

O objetivo desta tese foi compreender como o Programa de Aceleração do Crescimento (PAC), a partir dos investimentos no Polo Naval, solapa a coerência estruturada no município do Rio Grande, no Estado do Rio Grande do Sul. O estudo foi fundamentado na teoria marxista da dependência, especialmente nas discussões de Marini (2005) e mais recentemente de Osorio (2012a; 2012b), acerca da inserção subordinada das economias periféricas no mercado internacional, articulada com os mecanismos de acumulação do capital e de exploração do trabalho e na teoria da ordenação espaço-temporal de Harvey (2004) que permite a compreensão de como o capital produz espaço por meio do solapamento das estruturas que permitem a sua produção e reprodução. Sendo assim, a tese defendida neste trabalho é que o Programa de Aceleração do Crescimento, fundamentado em uma política neodesenvolvimentista, consolida o novo padrão de especialização produtiva no Brasil e altera a coerência estruturada das regiões que recebem investimentos por meio da criação de espaços dinâmicos de acumulação do capital e reafirma a dependência do país no sistema econômico mundial. Foi utilizada a abordagem dialético-materialista e a coleta de dados envolveu pesquisa documental, observação livre e entrevistas semiestruturadas com 34 pessoas pertencentes ao poder público, a sindicatos, a associações comunitárias e à comunidade em geral. A análise dos períodos econômico-produtivos pelos quais Rio Grande passou revela que a acumulação do capital na região é alicerçada na dependência externa e na segregação social. Em todos os períodos é evidente a presença do capital exógeno à cidade, fruto de ordenações espaço-temporais sistêmicas atraídas fundamentalmente pelo porto existente e de intervenções estatais na economia local. Esses processos geraram ciclos de imigração na cidade, uma vez que o capital necessita de força de trabalho para criar uma nova capacidade produtiva. Em todos os períodos também é notável a constatação de que os produtos produzidos no município eram destinados fundamentalmente para a exportação, o que o deixou ainda mais vulnerável a crises internacionais e causaram enormes prejuízos sociais quando entraram em declínio. Essas condições também são reproduzidas no último período econômico-produtivo analisado: a implementação do Polo Naval a partir dos investimentos do PAC. Este Programa, ao mesmo tempo em que norteia os planos estratégicos de desenvolvimento conduzidos pelo Governo Federal, é provocador do novo padrão exportador de especialização produtiva no Brasil e revisita sob uma nova roupagem os sinais da lógica da dependência, ao dar forma a uma contraditória composição entre autonomia nacional para o enfrentamento das desigualdades, com adequações ao sistema econômico mundial. Essas contradições, originadas nas disputas pelo projeto de nação, desenham as ações e prioridades de investimentos inscritos no Programa e evidenciam o modo de inserção do Brasil no sistema econômico mundial. Por estas razões, a teoria marxista da dependência mostra sua potência para interpretar uma política pública, configurada pelas leis de funcionamento do capital e pelo papel subsidiário que nele é destinado aos países periféricos da América Latina, e que o novo desenvolvimentismo não altera.
The aim of this thesis was to understand how the Growth Acceleration Program (PAC) from investments in Naval Pole, undermines the structured coherence in the city of Rio Grande, in Rio Grande do Sul. The study was based on marxist theory of dependency especially in discussions of Marini (2005) and more recently Osorio (2012a, 2012b), about the subordinate position of peripheral economies in the international market, combined with the mechanisms of accumulation of capital and labor exploitation and the theory of spatial-temporal ordering of Harvey (2004) that allows the understanding of how capital produces space through undermining the structures that enable its production and reproduction. Therefore, the argument of this thesis is that the Growth Acceleration Program, based on a new developmentalism policy, consolidates the new pattern of productive specialization in Brazil and alters the structured coherence of the regions that receive investments by creating dynamic spaces of capital accumulation and reaffirms the country‘s dependence on the world economic system. The dialectical-materialist approach were used and data collection involved documental research, free observation and semi-structured interviews with 34 people belonging to the government, labor unions, the community associations and the community in general. The analysis of the economic-productive periods for which Rio Grande went reveals that the accumulation of capital in the region is rooted in external dependence and social segregation. In all periods it is evident the presence of exogenous capital to the city, the result of ordinations spatiotemporal systemic attracted primarily by the existing port and government intervention on the local economy. These processes generate cycles of immigration in the city, once the capital needs of the workforce to create a new productive capacity. In all periods is also noteworthy to observe that the products produced in the city were intended primarily for export, which made him even more vulnerable to international crises and caused enormous social damage when went into decline. These conditions also reproduced in the final economic-productive period analyzed: the implementation of Naval Pole from the PAC investments. This program, while guiding the strategic development plans conducted by the Federal Government, is the provocative newcomer pattern of productive specialization in Brazil and revisits under a new guise signs of logical dependence, to form a composition contradictory between national autonomy for addressing inequalities, with adaptations to the world economic system. These contradictions, disputes arising in the project of nation, drawing the actions and priorities of investments included in the program and show the way of Brazil's insertion in the global economic system. For these reasons, the marxist theory of dependency shows its power to interpret a policy configured for the proper functioning of capital and the subsidiary role it is intended that the peripheral countries of Latin America, and the new developmentalism does not change.

碩士
國立臺灣大學
資訊網路與多媒體研究所
103
In recent years, structural knowledge base have attracted much attention in information retrieval and natural language processing. People rely on structural data to implement many applications that benefit the human life. With the growth of internet it is almost impossible for human editors to update all the world knowledge generated by people all around the world everyday to knowledge base. How to accelerate the construction of knowledge base becomes a critical issue known as knowledge base acceleration. Relation extraction technique can be the key part to accelerate the knowledge base construction progress and to extract the relation between entities. There are useful resources such as relation patterns which can denote the binary relation between two entities. However, there are few relation patterns resources available in Chinese information extraction. In this work, we present a Chinese relation patterns taxonomy for knowledge base acceleration. Each pattern in this taxonomy is semantically typed into YAGO properties and has its own confidence and entity types defined. We will describe the complete method to collect the Chinese corpora and to extract the Chinese relation patterns from them. Finally, we will examine the correctness of those patterns to evaluate the performance of the proposed pattern extraction method and analyze the errors occurred during the experiment. With the Chinese relation patterns taxonomy, many related works can be transferred from English to Chinese environments and further improve the usability and scale of Chinese knowledge bases.

碩士
國立臺灣大學
資訊網路與多媒體研究所
104
With the rapid development of the Internet in recent years, people can get infor-mation from it through different sources such as online news, social network, and fo-rums. A lot of information is created by people every day and some of them can be col-lected, comprehended, and turned into knowledge by human beings. Knowledge base is a way that people store those information with structural format. However, it’s hard to keep knowledge base up-to-date because of the wide gap between limited editors and numerous information of entities. Knowledge base acceleration is a critical issue which focus on accelerating the construction of knowledge base. In addition, relation patterns are useful for knowledge base acceleration. However, there are no resources available in languages beyond English. In this study, we present a workflow for building relation pattern extraction system with high coverage for knowledge base acceleration and knowledge base completion. Our properties is based on the properties in DBpedia knowledge base. We will discuss many details of our method including corpus pre-processing, instance retrieval, and pat-tern extraction. Finally, we evaluate our relation patterns by human annotators and dis-cuss features that may affect the performance of the relation patterns. With Chinese relation patterns, many related work can be utilized in Chinese by transferring from English environment to Chinese environment. Other languages may also use our method to build their own relation pattern resources.

碩士
國立臺灣師範大學
運動競技學系
104
Introduction：The population engaged in exercise increases annually. One of the main reasons for exercising is for promoting overall health; therefore, the accuracy of measuring physical activity is critical. In recent years, the electronic industry has been booming and wearable technology is getting increasingly mature. The accelerometer has made a big breakthrough in accuracy, volume and convenience and is now a necessity for recording physical activity. Although accelerometers are commonly used in measuring physical activity, the algorithms are quite diverse and not yet standardized. The moving speed is an index of exercise intensity but the correlations between the physical acceleration and moving speed are not yet clear. Purpose: The purpose of this study is to examine: 1) the correlations between the physical acceleration, moving speed and heart rate reserve (HRR%)； 2） a better algorithm for estimating physical acceleration at different walking and running speeds. Method: This study recruited 19 healthy males who were asked to wear a heart rate monitor and a tri-axis accelerometer to perform walking and running at different speeds. Each speed was tested for 3 minutes followed by a rest until the participants returned to their standing rest heart rate. A one-minute stable heart rate of each trial was gathered for data analysis. After being processed with different algorithms the acceleration values were analyzed with moving speeds and HRR% respectively. The Pearson product-moment correlation coefficient was used to examine the relationships between walking, running, the entire walking-running period, physical acceleration, moving speeds and HRR%. Result: In the walking, running, and the entire walking-running period, the peak values of resultant acceleration and the MAD values of anterior-posterior acceleration showed significant correlations with moving speeds (r=.829**、.514**、.836**； r=.833**、.637**、.780**) and HRR% (r=.771**、.517**、.856**； r= .837**、.651**、.787**). Conclusion: It is feasible to estimate the exercise intensity for walking, running and the entire walking-running period with the peak values of resultant acceleration and the MAD values of anterior-posterior acceleration. The results can be used as a foundation for the future application in measuring physical activity.

碩士
國立中興大學
土木工程學系所
102
Resource-Constrained Project Scheduling (RCPS) has become a very popular management technique in recent construction industries. RCSP not only satisfies the traditional CPM scheduling, requiring the precedent relationship between work items, but also provided the project control under resource constraints. Through the development of RCPS, resources such as labor, equipment, and material can be arranged to achieve the goal of project scheduling optimization. However, since the rise of the worker welfare consciousness, the upper limit of work hours per day, and work days per month need to be complied following government labor regulations. Thus RCPS must consider additional constraints come from the abovementioned regulations, making the modeling of RCSP more complex and difficult to be solved. This study proposed a Pattern Generation technology, integrating Integer Programming, to accelerate solving RCPS. Our results show that the efficiency of RCSP modeled through Pattern Generation technology is substantially increased, and make the solution of large scale field problems more feasible. Resource-Constrained Project Scheduling (RCPS) has become a very popular management technique in recent construction industries. RCSP not only satisfies the traditional CPM scheduling, requiring the precedent relationship between work items, but also provided the project control under resource constraints. Through the development of RCPS, resources such as labor, equipment, and material can be arranged to achieve the goal of project scheduling optimization. However, since the rise of the worker welfare consciousness, the upper limit of work hours per day, and work days per month need to be complied following government labor regulations. Thus RCPS must consider additional constraints come from the abovementioned regulations, making the modeling of RCSP more complex and difficult to be solved. This study proposed a Pattern Generation technology, integrating Integer Programming, to accelerate solving RCPS. Our results show that the efficiency of RCSP modeled through Pattern Generation technology is substantially increased, and make the solution of large scale field problems more feasible.

碩士
國立彰化師範大學
資訊工程學系
105
The advancement of high speed network results in the increasing number of patterns day by day in NIDS (Network Intrusion Detection System). The pattern matching process in NIDS takes about 70% of the whole processing time. The AC (Aho-Corasick) algorithm is unable to find the faulty time effectively. In view of this, much research has used the features of GPU (Graphics Processing Unit) to improve the repeated searching process of the AC algorithm. One of the methods, PFAC (Parallel Failureless Aho-Corasick) Algorithm, improves the above problem effectively. This main idea of this algorithm is to remove backward branches in the FSM (Finite State Machine) in the original AC Algorithm created FSM. Moreover, it does not produce boundary detection. However, we found that this algorithm has a serious load imbalance problem, after analyzing it. Therefore, this paper proposes a two-phase where a threshold is predefined to divide the two phase. According to the experimented results, our approach can have up to 50% performance improvement than the PFAC algorithm.

碩士
國立成功大學
資訊工程學系碩博士班
101
Pattern matching plays an important role in Network Intrusion Detection System (NIDS) which is widely used in protecting computers from malicious attacks. With the rapid expansion of the network, the traditional single-core single-threaded CPU based algorithms have been unable to load demand, so it is necessary to accelerate pattern matching algorithm through hardware. GPU is a low cost and high scalability choice. With the development of Compute Unified Device Architecture (CUDA) technique by Nvdia, CUDA can promote the programmability of GPU program and let the role of GPU not only in executing graphical computing, but support the Single Instruction Multiple Thread (SIMT) big data computation. It can accelerate the speed of pattern matching effectively. In this thesis, we propose Multi-Byte Multi-Thread AC (MBMTAC) algorithm which is based on the famous Aho-Corasick (AC) algorithm and the Parallel Failureless-AC (PFAC) algorithm that improves the AC algorithm on GPU. (1) Base on the concept that let each byte of input stream as a start position and assign one thread for a start position to execute pattern matching, we extend it to matching multiple characters at once. The offset of each pattern does not need to be considered when establishing the data structure. Therefore, the problem of large amounts of additional states and transitions on CPU is solved. (2) Based on the concept, we do not need to consider the backward transitions, self-loops and the failure functions. Hence, the transition table becomes very sparse. We use a perfect hash scheme to compress the transition table. The memory usage of MBMTAC is only 0.19% of the PFAC needs. (3) By utilizing the characteristic of CUDA, we optimize the MBMTAC data structure so that the throughput can reach 33.4 Gb/s.

碩士
國立屏東科技大學
生物機電工程系所
97
This research uses three kinds of High-Power LED：UV light、Blue light and Red light. The purposes in the research are observing the changes of electric and optical characteristic in different wavelength by using high temperature acceleration aging test for a long time. And also uses optical simulation software TracePro to analyze the effects of lens structure and material aging on the radiation patterns. The reasons of optical power degradation are chip degradation and the decline of seal material. Yellowing and carbonization of lens and silicon gel, the bubbles and deformation on the surface of lens, and the degradation of chip making the increase of inverse current cause LED degradation.This research comp the ineffective model of different wavelength LED. The UV device of short-wave will speed up the seal material degrade; therefore the life time of short-wave is the shortest in the three wavelength. In the beginning, the device of blue light optical power decreases because of chip degradation. But silicon gel carbonization makes optical power decline obviously. The seal material of red light device becomes a little yellowing in high temperature environment, and its optical power decreases slightly. The life time of red light is the longest of all. The results show that after the deformation of outer lens structure and the aging of the seal material (PC and silicon gel), the observable rates increase 28.2% and 4.7%. the structure and material influence the light and optical power decreases about 30% between -52°~-90°, but optical power increases about 21% between -5°~-50°.

碩士
國立中山大學
光電工程研究所
96
The technology of high-power light-emitting diodes (LEDs) is mature. In this study, the LEDs were aged through high temperature to understand the packaging failure mechanisms under lasting lighting. The Tracepro optics simulation software is also applied to analyze the radiation pattern decay caused by curvature variation and material aging. In the experiment procedure, the LEDs of three manufactures were placed in aging ovens in three different temperatures set at 95℃, 85℃, and 65℃. The experiment results showed that the LED failures were the lens and silicon gel yellowing, silicon gel carbonization, lens slightly chapping, silicon cracks, bubbles generated between lens and silicon gel, and chip damage induced reverse current increase during the aging experiment. We observed the failure mechanism in reduction of optical power at 90%, 80%, 70%, 60%, 50% original power. Furthermore, the failure mechanisms of different from the manufactures is also compared in this work. The LED packages have long lifetime without silicon gel carbonization. In simulation study, a gap of lens deformation showed a reduction at -60°~-40° and 40°~60° of the radiation pattern. The LED modules also showed that the silicon gel and PC absorption coefficient increased. Material absorption coefficient increased more in the short wavelength showed 149mcd at ±60° of the radiation pattern.

碩士
國立交通大學
電子工程學系
85
In this thesis, we realize the hardware design and circuit simulation of a zero delay fault simulation hardware accelerator for the gate level synchronoussequential circuits. A parallel sequence fault simulation algorithm is used toachieve parallel-pattern fault simulation, and multiple processing elementsare used to achieve the parallel-fault fault simulation. As to detailed architectural and hardware design, we adopt several specific methodologiesand design approaches to fulfill the hardware fault simulation andaccelerate its performance. The proposed architecture is implemented inVerilog hardware description language and targeted for Compass 0.6u cell library. When the accelerator being configured as single PE, its performanceruns at an average from 1.95E+07 G*V/s to 8.22E+07 G* V/s. It's about 20 to 80times of software simulators. A speedup of 6.96 in average can be reached ifit is configured as 16 PEs.

Doutoramento em Motricidade Humana na especialidade de Biomecânica
This thesis aimed to provide a better understanding on the determinant factors for falling in Portuguese older adults, with a special emphasis on the biomechanical changes in gait patterns associated with the functional fitness decline in this population. Our methodological approach to this problem encompassed two different levels of analysis: in the first part two epidemiological studies were conducted in order to establish the determinant factors for falling within the Portuguese older adults; in the second part three laboratory-based studies were performed in order to determine the influence of functional fitness levels on elderly gait patterns. Falls were shown to result from the interaction of many risk factors. Within these, gender, functional fitness level and health parameters were found to be the strongest fall determinants. Interestingly, age was not a determinant factor for falling, even within very old individuals (≥75 years or ≥80 years). Therefore, in the subsequent studies, the gait patterns of a subgroup of older adults, who had participated in the epidemiological studies, were characterized according with their functional fitness levels. The results showed that older subjects with a lower functional fitness level score, consistently re-distribute lower limb joint moments while performing different locomotor tasks (walking, stair ascent and stair descent). Because the success of physical activity interventions aiming at falls and disability prevention is dependent on subgroup characterization, these biomechanical gait pattern changes may yield important information for the health and exercise professionals working with the elderly.
RESUMO: A presente dissertação objetiva o aprofundamento do conhecimento sobre os determinantes das quedas na população idosa portuguesa, com especial enfoque nas alterações biomecânicas nos padrões de marcha associadas ao declínio funcional característico desta população. A abordagem metodológica preconizada para a análise do problema compreende duas fases complementares: uma primeira fase, que englobou dois estudos epidemiológicos com o objetivo de estabelecer os fatores determinantes de quedas na população idosa portuguesa; uma segunda fase, onde foram considerados três estudos experimentais (laboratoriais), com o propósito de determinar a influência de diferentes níveis de aptidão funcional nos padrões de marcha desta população. Os resultados demonstraram que as quedas resultam da interação de diversos fatores de risco, destacando-se os seguintes: género, parâmetros de aptidão funcional e de saúde. De relevar que o fenómeno de queda se revelou independente da idade, mesmo quando analisada a sua associação com os fatores determinantes em grupos etários mais avançados (≥75 e ≥80 anos). Neste sentido, nos estudos subsequentes, foram analisados os padrões de marcha de subgrupos de idosos recrutados do grupo de participantes dos estudos anteriores e estratificados em função do seu nível de aptidão funcional. Observou-se então que os idosos com baixos níveis de aptidão funcional adotavam estratégias consistentes de redistribuição dos momentos de força articulares dos membros inferiores, aquando da execução de diferentes tarefas locomotoras (marcha, subir e descer escadas). Considerando o sucesso demonstrado das intervenções sustentadas em programas de atividade física para a prevenção de quedas e incapacidade, as alterações biomecânicas dos padrões de marcha observadas poderão constituir um importante suporte informacional para os profissionais de saúde e exercício que trabalham com a população idosa.
FCT - Fundação para a Ciência e a Tecnologia