京都大学 / 0048 / 新制・課程博士 / 博士(工学) / 甲第22047号 / 工博第4628号 / 新制||工||1722(附属図書館) / 京都大学大学院工学研究科社会基盤工学専攻 / (主査)教授 後藤 仁志, 教授 細田 尚, 准教授 KHAYYER,Abbas / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DGAM

京都大学 / 0048 / 新制・課程博士 / 博士(情報学) / 甲第22580号 / 情博第717号 / 新制||情||123(附属図書館) / 京都大学大学院情報学研究科社会情報学専攻 / (主査)教授 吉川 正俊, 教授 黒橋 禎夫, 教授 田島 敬史, 教授 田中 克己(京都大学 名誉教授) / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

京都大学 / 新制・課程博士 / 博士(情報学) / 甲第23548号 / 情博第778号 / 新制||情||132(附属図書館) / 京都大学大学院情報学研究科通信情報システム専攻 / (主査)教授 湊 真一, 教授 山本 章博, 准教授 川原 純 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

Trots att Sverige anses vara ett jämställt land finns det stora löneskillnader mellan kvinnor och män. Dessa löneskillnader tycks vara extra stora inom bank-och finansbranschen och kan till viss del förklaras av frånvaroperioder från arbetet då kvinnor tenderar att vara deltidsarbetande, föräldralediga, sjukfrånvarande och frånvarande på grund av VAB i större utsträckning än män. Denna studie syftar därmed till att undersöka löneskillnader på Handelsbanken Capital Markets ur ett jämställdhetsperspektiv. Kvinnors och mäns löner undersöks med fokus på om det finns ett samband mellan olika sorters frånvaro från arbetet och löneutvecklingen. Studiens teoretiska ramverk baseras på teorierna om statistisk och implicit diskriminering samt teorin om signaleffekt. Dessa utgår från att kvinnor och män drabbas av lönediskriminering på grund av arbetsgivarens antaganden, statistiska erfarenheter och förutfattade meningar om bland annat genusnormer. Av tidigare forskning framgår att mäns löner påverkas mer negativt än kvinnors löner av frånvaro i olika former. Denna studie baseras på Handelsbanken Capital Markets löneregisterdata från 2017, 2018 och 2019. Urvalet består av 691 individer som varit anställda och har haft en månadslön under hela tidsperioden. Resultaten visar att män som varit deltidsarbetande, föräldralediga, sjukfrånvarande eller frånvarande på grund av vård av barn tenderar att ha sämre löneutveckling än kvinnor som varit frånvarande av samma anledning. Däremot är det enbart resultatet för sjukfrånvaro som är statistiskt signifikant.

The purpose of this thesis is to solve unsteady two-dimensional compressible Navier-Stokes equations for a moving mesh using implicit explicit (IMEX) Runge- Kutta scheme. The moving mesh is implemented in the equations using Arbitrary Lagrangian Eulerian (ALE) formulation. The inviscid part of the equation is explicitly solved using second-order Godunov method, whereas the viscous part is calculated implicitly. We simulate subsonic compressible flow over static NACA-0012 airfoil at different angle of attacks. Finally, the moving mesh is examined via oscillating the airfoil between angle of attack = 0 and = 20 harmonically. It is observed that the numerical solution matches the experimental and numerical results in the literature to within 20%.

Although much research has been done on how well working memory predicts processing of consciously activated information, research on the possible influences of working memory on automatically activated information is scarce (Barrett, Tugade, & Engle, 2004). Working memory capacity (WMC) may be related to how much information is activated automatically by either aiding ease of access to relevant information or by its role in inhibiting irrelevant thoughts and information (i.e., noise). The purpose of the present study was to examine the contribution of individual differences in WMC on implicit and explicit processes in cued recall. Participants studied target words and recall was cued by associatively related words. Target connectivity was varied in Experiment 1 and target set size was varied in Experiment 2. The cued recall memory test was conducted after various retention intervals (0, 10 and 20 mins). In addition, memory span of all participants was measured with both operation and counting span tasks. Finally, all participants studied a second list of words under divided attention instructions. The present experiments examined 1) the influence of retention interval on cued recall performance, 2) the influence of individual differences in WMC on cued recall after various retention intervals and 3) the role of WMC and divided attention on implicitly activated knowledge (i.e., connectivity and set size effects). The findings revealed that working memory is related to intentional (explicit) types of processes, but not related to implicit processes outside of a person's awareness. WMC also interacted with retention interval. This finding is compatible with an attentional interpretation of WMC that assumes the high span advantage is apparent only when there is interference. Surprisingly, low span participants tended to outperform high span individuals on an immediate test. These findings are explained by differences in maintenance of information and rehearsal, and retrieval strategies.

Un défi pour les systèmes de recherche basée sur le contenu réside dans la nécessité d’avoir une base annotée. Cette thèse propose un système d’annotation d’images interactif par le regard afin d’alléger la tâche d’annotation. Le but est de classer un petit ensemble d’images en fonction d’une catégorie cible (classification binaire) pour classer un grand ensemble d’images. Parmi les caractéristiques du regard pointées comme informatives sur l’intention des utilisateurs, nous avons élaboré un estimateur d’intention par le regard, calculable en temps réel, indépendant de l’utilisateur et de la catégorie cible. Cette annotation implicite est meilleure qu’une annotation aléatoire mais reste incertaine. Dans une deuxième partie, les images ainsi annotées sont utilisées pour classifier un plus grand ensemble d’images avec un algorithme prenant en compte l’incertitude des labels : P-SVM combinant classification et régression. Nous avons déterminé parmi différentes stratégies un critère de pertinence pour discriminer les labels les plus fiables, utilisés pour la classification, des labels les plus incertains, utilisés pour la régression. La précision du P-SVM est évaluée dans différents contextes et peut atteindre les performances d’un algorithme de classification standard entraîné avec les labels certains. Ces évaluations ont tout d’abord été menées sur un benchmark standard pour se comparer à l’état de l’art, et dans un second temps, sur une base d’images de nourriture. / One daunting challenge of Content Based Image Retrieval systems is the requirement of annotated databases. To limit the burden of annotation, this thesis proposes a system of image annotation based on gaze data. The purpose is to classify a small set of images according to a target category (binary classification) in order to classify a set of unseen images. First, we have designed a protocol based on visual preference paradigm in order to collect gaze data from different groups of participants during a category identification task. Among the gaze features known to be informative about the intentions of the participants, we have determined a Gaze-Based Intention Estimator (GBIE), computable in real-time; independent from both the participant and the target category. This implicit annotation is better than random annotation but is inherently uncertain. In a second part, the images annotated by the GBIE from the participants’ gaze data are used to classify a bigger set of images with an algorithm that handles label uncertainty: P-SM combining classification and regression SVM. We have determined among different strategies a criterion of relevance in order to discriminate the most reliable labels, involved in the classification part, from the most uncertain labels, involved in the regression part. The average accuracy of P-SVM is evaluated in different contexts and can compete with the performances of standard classification algorithm trained with true-class labels. These evaluations were first conducted on a standard benchmark for comparing with state-of-the-art results and later conducted on food image dataset.

Millimeter wave (mmWave) spectrum above 30 GHz offers us an opportunity to pursue high-data-rate transmission using a channel bandwidth up to several gigahertz. To provide reliable link quality in mmWave frequency bands, hybrid analog-digital beamforming plays a crucial role in overcoming severe path loss and, meanwhile, satisfies the demand for low-power-consumption radio frequency (RF) devices. Implementing hybrid beamforming based on available channel state information (CSI) is a common solution in the hybrid beamforming literature. However, many reference methods underestimate the computational complexity of channel estimation for large antenna arrays or subsequent steps, such as the singular value decomposition of a channel matrix. To this end, we present a low-complexity scheme that exploits implicit channel knowledge to facilitate the design of hybrid beamforming for frequency-selective fading channels. We start from the study of channel estimation using the orthogonal matching pursuit (OMP) algorithm and realize that the problems of channel estimation and analog beam selection are equivalent if the candidates for analog beamforming vectors in the codebooks are mutually orthogonal. This implies that when orthogonal codebooks are employed, the observations used in channel estimation for large antenna arrays can be used to implement hybrid beamforming directly. The above-mentioned observations can be regarded as \textbf{implicit CSI}; they are coupling coefficients of all possible pairs of analog beamforming vectors on both sides of the channel. The idea of using implicit CSI to implement hybrid beamforming is further extended to the cases of non-orthogonal codebooks. Instead of calculating the mutual information between the transmitter and receiver, we focus on small-size coupling matrices between beam patterns selected by using appropriate key parameters as performance indicators. Therefore, the proposed hybrid beamforming method becomes much simpler: it amounts to collecting different sets of large-power coupling coefficients to construct multiple alternatives to an effective channel matrix. Then, the set yielding the largest Frobenius norm (or the largest absolute value of the determinant) of the effective channel provides the solution to the hybrid beamforming problem. The proposed hybrid beamforming approach clearly shows that the performance of hybrid beamforming is dominated by the quality of the coupling coefficients. Considering a fixed-length training sequence, we exploit mmWave channels' sparsity shown in the delay and angular domains to refine the quality of the coupling coefficients as well as to improve the hybrid beamforming performance.

This thesis focuses on solutions of reactive transport problems in porous media. The principle mechanisms of flow and reactive mass transport in porous media are investigated. Global implicit approach (GIA), where transport and reaction are fully coupled, and sequential noniterative approach (SNIA) are implemented into the software OpenGeoSys (OGS6) to couple chemical reaction and mass transport. The reduction scheme proposed by Kräutle is used in GIA to reduce the number of coupled nonlinear differential equations. The reduction scheme takes linear combinations within mobile species and immobile species and effectively separates the reaction-independent linear differential equations from coupled nonlinear ones (i.e. reducing the number of primary variables in the nonlinear system). A chemical solver is implemented using semi-smooth Newton iteration which employs complementarity condition to solve for equilibrium mineral reactions. The results of three benchmarks are used for code verification. Based on the solutions of these benchmarks, it is shown that GIA with the reduction scheme is faster (ca. 6.7 times) than SNIA in simulating homogeneous equilibrium reactions and (ca. 24 times) in simulating kinetic reaction. In simulating heterogeneous equilibrium mineral reactions, SNIA outperforms GIA with the reduction scheme by 4.7 times.:Declaration of Authorship iii Acknowledgements iv Abstract v List of Figures viii Symbols ix 1 Introduction 1 1.1 State of the Art 1 1.2 Thesis Objectives 3 1.3 Thesis Outline 4 2 Mathematical Models 5 2.1 Introduction 5 2.2 Mass Balance Equations 5 2.2.1 Groundwater Flow 6 2.2.2 Mass Transport 7 2.2.3 Chemical Reaction 8 2.2.3.1 Equilibrium Reaction 8 2.2.3.2 Kinetic Reaction 10 2.3 Reactive Mass Transport 10 2.4 Initial and Boundary Conditions 11 3 Numerical Solutions 12 3.1 Introduction 12 3.2 Coupling Schemes 12 3.2.1 Operator Splitting 13 3.2.2 Global Implicit 13 3.2.2.1 Standard Reduction Schemes 14 3.2.2.2 Kräutle’s Reduction Scheme 14 3.2.2.3 Local Chemical Solver 21 3.3 Space and Time Discretization 23 3.3.1 Finite Element Method 23 3.3.2 Time Discretization 25 3.3.3 Jacobian Matrix 26 3.4 Code Implementation 29 4 Benchmarks 30 4.1 Introduction 30 4.2 Cation Exchange 30 4.3 Dissolution and Precipitation 32 4.4 Mixing Controlled Biodegradation 33 5 Conclusions and Outlooks 38 5.1 Conclusions 38 5.2 Outlooks 39 / Diese Arbeit konzentriert sich auf die numerische Berechnung reaktiver Transportprobleme in porösen Medien. Es werden prinzipielle Mechanismen von Fluidströmung und reaktive Stofftransport in porösen Medien untersucht. Um chemische Reaktionen und Stofftransport zu koppeln, wurden die Ansätze Global Implicit Approach (GIA) sowie Sequential Non-Iterative Approach (SNIA) in die Software OpenGeoSys (OGS6) implementiert. Das von Kräutle vorgeschlagene Reduzierungsschema wird in GIA verwendet, um die Anzahl der gekoppelten nichtlinearen Differentialgleichungen zu reduzieren. Das Reduzierungsschema verwendet Linearkombinationen von mobilen und immobile Spezies und trennt die reaktionsunabhngigen linearen Differentialgleichungen von den gekoppelten nichtlinearen Gleichungen (dh Verringerung der Anzahl der Primärvariablen des nicht-linearen Gleichungssystems). Um die Gleichgewichtsreaktionen der Mineralien zu berechnen, wurde ein chemischer Gleichungslaser auf Basis von ”semi-smooth Newton-Iterations” implementiert. Ergebnisse von drei Benchmarks wurden zur Code-Verifikation verwendet. Diese Ergebnisse zeigen, dass die Simulation homogener Equilibriumreaktionen mit GIA 6,7 mal schneller und bei kinetischen Reaktionen 24 mal schneller als SNIA sind. Bei Simulationen heterogener Equilibriumreaktionen ist SNIA 4,7 mal schneller als der GIA Ansatz.:Declaration of Authorship iii Acknowledgements iv Abstract v List of Figures viii Symbols ix 1 Introduction 1 1.1 State of the Art 1 1.2 Thesis Objectives 3 1.3 Thesis Outline 4 2 Mathematical Models 5 2.1 Introduction 5 2.2 Mass Balance Equations 5 2.2.1 Groundwater Flow 6 2.2.2 Mass Transport 7 2.2.3 Chemical Reaction 8 2.2.3.1 Equilibrium Reaction 8 2.2.3.2 Kinetic Reaction 10 2.3 Reactive Mass Transport 10 2.4 Initial and Boundary Conditions 11 3 Numerical Solutions 12 3.1 Introduction 12 3.2 Coupling Schemes 12 3.2.1 Operator Splitting 13 3.2.2 Global Implicit 13 3.2.2.1 Standard Reduction Schemes 14 3.2.2.2 Kräutle’s Reduction Scheme 14 3.2.2.3 Local Chemical Solver 21 3.3 Space and Time Discretization 23 3.3.1 Finite Element Method 23 3.3.2 Time Discretization 25 3.3.3 Jacobian Matrix 26 3.4 Code Implementation 29 4 Benchmarks 30 4.1 Introduction 30 4.2 Cation Exchange 30 4.3 Dissolution and Precipitation 32 4.4 Mixing Controlled Biodegradation 33 5 Conclusions and Outlooks 38 5.1 Conclusions 38 5.2 Outlooks 39

The research aims to focus on Recurrent Neural Networks (RNN) and its application to the session-aware recommendations empowered by implicit user feedback and content-based metadata. To investigate the promising architecture of RNN, we implement seven different models utilizing various types of implicit feedback and content information. Our results showed that using RNN with complex implicit feedback increases the next-item prediction comparing the baseline models like Cosine Similarity, Doc2Vec, and Item2Vec.