No description available.

京都大学 / 新制・課程博士 / 博士(情報学) / 甲第24821号 / 情博第837号 / 新制||情||140(附属図書館) / 京都大学大学院情報学研究科知能情報学専攻 / (主査)教授 熊田 孝恒, 教授 西田 眞也, 教授 内田 由紀子, 准教授 中島 亮一 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

<p>Anticipating the occurrence of a future event is an ability that helps people prepare for various daily activities. This preparation is regarded as a non-specific process because it is initiated by a warning signal that does not contain specific information about the critical event. Previous research reported that the intertrial repetition of a stimulus-response event in a choice-reaction task shortened the reaction time more at the short foreperiod (interval between the end of the warning signal and onset of the target stimulus). I conducted four experiments to investigate whether the interaction was due to the event sequence effect being overridden by preparation processes (“overriding” hypothesis) or the quick-decaying characteristic of the event sequence effect itself (“quick-decay” hypothesis). Experiments 1 and 2 manipulated the relative magnitudes of the preparation effect by changing how foreperiods were distributed within a trial block. The results showed similar asymmetric event sequence effects, which indicated that whether preparation was better at the short or long foreperiod did not affect the event-based modulation. Experiment 3 manipulated the temporal distance between two consecutive stimulus-response events across trial blocks and found that the asymmetric event-based modulation on preparation was diminished by a long enough inter-trial interval. The final experiment compared alerting trials with no-alerting trials and found an asymmetric event-based modulation caused by the absence of repetition benefit in a certain context (an alerting trial preceded by a no-alerting trial). Therefore, the event sequence effect is not directly related to “nonspecific preparation”, but this event-specific component could be embedded in the measurement of preparation in some scenarios, which could lead to misinterpretation of the preparation effect itself. This finding clarifies the mechanism underlying the interaction between preparation and event sequence. The conclusion also questions the validity of the current measures of nonspecific preparation, including temporal preparation and phasic alertness.</p>

<p dir="ltr">In my thesis, I propose a Graph Regularized-Attention-Based Diffusion Transformer (GRAD-T) model, which uses kernel temporal attention and a regularized sparse graph method to analyze model general diffusion processes over networks. The proposed model uses the spatiotemporal nature of data generated from diffusion processes over networks to examine phenomena that vary across different locations and time, such as disease outbreaks, climate patterns, ecological changes, information flows, news contagion, transportation flows or information and sentiment contagion over social networks. The kernel attention models the temporal dependence of diffusion processes within locations, and the regularized spatial attention mechanism accounts for the spatial diffusion process. The proposed regularization using a combination of penalized matrix estimation and a resampling approach helps in modeling high-dimensional data from large graphical networks, and identify the dominant diffusion pathways. I use the model to predict how emotions spread across sparse networks. I applied the model to a unique dataset of COVID-19 tweets that I curated, spanning April to July 2020 across various U.S. locations. I used model parameters (attention measures) to create indices for comparing emotion diffusion potential within and between nodes. Our findings show that negative emotions like fear, anger, and disgust demonstrate substantial potential for temporal and spatial diffusion. Using the dataset and the proposed method we demonstrate that different types of emotions exhibit different patters of temporal and spatial diffusion. I show that the proposed model improves prediction accuracy of emotion diffusion over social medial networks over standard models such as LSTM and CNN methods. Our key contribution is the regularized graph transformer using a penalty and a resampling approach to enhance the robustness, interpretability, and scalability of sparse graph learning.</p>