No description available.

No description available.

Data profiling is the process of analyzing data and producing an output with statistical summaries. The size of data rapidly increases and it is more difficult to process all data in a reasonable time. All data can not be stored in RAM memory, so it is not possible to run exact single-pass algorithms without using slower computer storage. The diploma thesis focuses on the implementation, comparison, and selection of suitable algorithms for data profiling of large input data. Usage of approximate algorithms brings a possibility to limit mem- ory for computation, do the whole process in RAM memory and the duration of data profiling should be reduced. The tool can compute frequency analysis, cardinality, quantiles, histograms, and other single-column statistics in a short time with a relative error lower than one percent.

The operation of the Solitaire algorithm is based on a deck of cards, including two different jokers. We use the computer to simulate the Solitaire algorithm and analyze the key stream which is generated by the algorithm.We mainly analyze the maximum cycle length of the key stream and doNIST test.

The purpose of this dissertation is to demonstrate the ability to design health monitoring systems from a systematic perspective and how, with proper sensor and actuator placement, damage occurring in a structure can be detected and tracked. To this end, a design optimization was performed to determine the best locations to excite the structure and to collect data while using the minimum number of sensors. The type of sensors used in this design optimization was uni-axis accelerometers. It should be noted that the design techniques presented here are not limited to accelerometers. Instead, they allow for any type of sensor (thermal, strain, electromagnetic, etc.) and will find the optimal locations with respect to defined objective functions (sensitivity, cost, etc.). The use of model-based optimization techniques for the design of the monitoring system is driven by the desire to obtain the best performance possible from the system given what is known about the system prior to implementation. The use of a model is more systematic than human judgment and is able to take far more into account by using information about the dynamical response of a system than even an experienced structural engineer. It is understood in the context of structural modeling that no model is 100\% accurate and that any designs produced using model-based techniques should be tolerant to modeling errors. Demonstrations performed in the past have shown that poorly placed sensors can be very insensitive to damage development. To perform the optimization, a multi-objective genetic algorithm (GA) was employed. The objectives of the optimization were to be highly sensitive to damage occurring in potential “hot spots” while also maintaining the ability to detect damage occurring elsewhere in the structure and maintaining robustness to modeling errors. Two other objectives were to minimize the number of sensors and actuators used. The optimization only considered placing accelerometers, but it could have considered different type of sensors (i.e. strain, magneto-restrictive) or any combination thereof.

Although past research finds that auditors support data analytics and artificial intelligence to enhance audit quality in their daily work, in reality, only a small number of audit firms, who innovated and invested in the two sophisticated technologies, utilize it in their auditing process. This paper analyzes three factors, including three individual theories, that may influence the adoption of data analytics and artificial intelligence in auditing: regulation (Institutional theory: explaining the catch-22 between the auditors and policymakers), knowledge barrier (Technology acceptance model's theory: explore the concept of ease of use), and people (algorithm aversion: a phenomenon that auditors believe in human decision makers more than technology). Among the three barriers, this paper focuses more on the people factor, which firms can start to overcome early. Past research has shown the existence of algorithm aversion in audit, so it is important to identify ways to decrease algorithm aversion. This study conducted a survey with four attributes: transparency-efficiency-trade-off, positive exposure, imperfect algorithm, and company's training. The study results shows that transparency-efficiency-trade-off can be a potential solution for decreasing algorithm aversion. When auditor firms implement transparency-efficiency-trade-off in their company training, auditors may give more trust to the technologies. The trust may lead to the increase of data analytics and artificial intelligence in audit.

Clustering is the process of finding underlying group structure in data. As the scale of data collection continues to grow, this “big data” phenomenon results in more complex data structures. These data structures are not always compatible with traditional clustering methods, making their use problematic. This thesis presents methodology for analyzing samples of four-way and higher data, examples of these more complex data types. These data structures consist of samples of continuous data arranged in multidimensional arrays. A large emphasis is placed on clustering this data using mixture models that leverage tensor-variate distributions to model the data. Parameter estimation for all these methods are based on the expectation-maximization algorithm. Both simulated and real data are used for illustration. / Thesis / Doctor of Science (PhD)

京都大学 / 新制・課程博士 / 博士(理学) / 甲第24388号 / 理博第4887号 / 新制||理||1699(附属図書館) / 京都大学大学院理学研究科数学・数理解析専攻 / (主査)教授 牧野 和久, 教授 並河 良典, 教授 長谷川 真人 / 学位規則第4条第1項該当 / Doctor of Science / Kyoto University / DGAM

No description available.

No description available.