187037 research outputs found
Sort by
웹툰 불법 사이트 이용자들의 공식 플랫폼으로의 전환 의도에 대한 연구
No full text학위논문(석사) - 한국과학기술원 : 기술경영학부, 2025.2,[iii, 32 p. :]Originated from South Korea in the 2010’s, Webtoon is an increasingly popular form of cartoon distributed and read online. Despite its popularity worldwide, Webtoon market is heavily affected by illegal distribution. Therefore, the aim of this study is to identify factors affecting consumers’ illegal Webtoon site usage and to analyze how inertial use of illegal sites affect their intention to switch to official platforms. By conducting an online survey (n=213), this study finds that a user’s novelty-seeking propensity positively affects their status quo inertia to use illegal sites, and that subjective norms negatively affects inertia, while also positively affecting the user’s intention to use official platforms. Moreover, based on status quo bias theory and coping theory, status quo inertia has been found to have both direct and indirect effects on the intention to use official platforms, but its downward bias on the opportunity appraisal has not been confirmed. Lastly, the opportunity appraisal of official platforms significantly affected the user’s intention to use them.한국과학기술원 :기술경영학부
자연에서 영감을 얻은 산소 라디칼 제거형 첨가제 라이코펜 도입을 통한 리튬 이온 전지의 열 안전성 향상
No full text학위논문(석사) - 한국과학기술원 : 생명화학공학과, 2025.2,[ii, 24 p. :]For a green environment, substituting fossil fuels with renewable energy sources is inevitable. One of the preferred solutions is lithium-ion battery, commonly used in portable electric devices and electric vehicles. However, lithium-ion batteries are suffering from fatal safety issues such as fire accidents and battery pack explosions caused by mechanical, electrical and thermal mistreatment of cells. Especially, gases produced by the degradation of Ni-rich cathode material and decomposition of liquid electrolytes has been interacting with each other, resulting in persistent exothermic reactions until thermal runaway occur. In this study, a novel nature-inspired antioxidant additive lycopene is introduced for enhanced thermal stability of lithium-ion battery at elevated temperature. Assessment of super oxide radical quenching ability of lycopene containing electrolyte along with its gas scavenging mechanism and full cell cycle performance has been thoroughly explored.한국과학기술원 :생명화학공학과
막 유화법으로 제조된 에멀젼 템플릿을 통한 단분산 중공형 실리카의 대량 생산에 관한 연구
No full text학위논문(석사) - 한국과학기술원 : 생명화학공학과, 2025.2,[ii, 29 p. :]Hollow silica has garnered significant attention for its applications in various fields such as drug delivery, energy storage devices, and catalysts, owing to its unique properties provided by the hollow structure and silica matrix. Conventionally, hard templates such as polymer particles or soft templates like emulsions, micelles, vesicles, and gas bubbles are employed as sacrificial templates. While hard templates enable precise control over morphology, size, and size distribution, their synthesis and removal processes are complex and hinder scalability. Conversely, soft templates offer greater scalability but suffer from poor control over morphology and uniformity. Balancing the advantages of both methods—achieving high uniformity alongside scalability—remains a significant challenge.
In this study, we attempted to address this trade-off by producing monodisperse hollow silica microspheres using an emulsion generated via a membrane emulsification technique as a template. For confined sol-gel reaction of silica precursor at the interface, CTAB was added to the water phase. In result, robust silica capsules was fabricated instead of solid silica particles. The effect of reaction time, surfactant concentration, reaction temperature, and TEOS concentration on silica capsule morphology was investigated by optical microscopy and SEM analysis. Shell thickness of the capsules could be tuned from 32.4 nm to 324.8 nm by controlling the reaction time. When CTAB-containing water phase was used, the membrane emulsification method did not proceed successfully, resulting in polydisperse and big droplets. To avoid the use of CTAB, APTES was introduced as a comonomer with TEOS to synthesize capsules. It was shown that polydisperse capsules compared to the initial monodisperse emulsion were synthesized, due to decreased stability of emulsion after precursor mixture was added. Therefore, more precise control over the sol-gel reaction rate should be studied to form stable silica capsules before the emulsion stability deteriorates.한국과학기술원 :생명화학공학과
커널 기반 고차 열화 모델링을 통한 초해상화 성능 향상
No full text학위논문(석사) - 한국과학기술원 : 전산학부, 2025.2,[iii, 21 p :]In this paper, we propose a novel blind super-resolution method that integrates a kernel-based degradation model with Discrete Cosine Transform (DCT) loss to improve image restoration performance. Existing super-resolution techniques often fail to account for complex degradation processes. To address this, our approach uses kernel estimation based on image patch distributions to more effectively simulate the complex degradations present in real-world images. Additionally, the DCT-based loss function plays a key role in preserving the overall frequency information of the image. Experiments conducted on benchmark datasets and the DPED dataset demonstrate that the proposed method exhibits enhanced robustness in handling complex image degradations, and shows promising performance for high-quality image restoration in real-world applications.한국과학기술원 :전산학부
효율적인 노드 분류를 위한 이종 그래프 신경망에서 경량 학생 모델로의 지식 증류
No full text학위논문(석사) - 한국과학기술원 : 전산학부, 2025.2,[iv, 26 p. :]We address the challenge of Knowledge Distillation (KD) from teacher Heterogeneous Graph Neural Networks (HGNNs) to student Multi-Layer Perceptrons (MLPs) for semi-supervised node classification tasks. This work extends the current research on graph knowledge distillation, which mainly focuses on homogeneous graphs, to heterogeneous graphs, commonly used in real-world industrial applications. To reflect the rich semantics and diverse connections of heterogeneous graphs, we propose a new framework that leverages semantic high-order Metapath2vec features and global structure-aware DeepWalk features extracted from input graphs. The feature combination strategy of weighted combination and fusion is incorporated into the knowledge distillation process to balance the logits learned by the teacher model and the features input to the student model. Experimental results on three heterogeneous graph datasets using three HGNNs as teacher models show that the student model, trained with additional node features, achieves an average performance improvement of 1.94% while being approximately 328x faster than the teacher model. Moreover, it demonstrates a 16.67% performance improvement compared to a baseline MLP that does not use KD or the additional semantic and structural features.한국과학기술원 :전산학부
The impact of natural disasters on firm search orientation: Evidence from a natural experiment
No full textThis study examines the influence of natural disasters on firm search orientation in technology development. While prior research has extensively explored the environmental antecedents of exploration and exploitation, the potential impact of these environmental antecedents on decision-makers' risk preference has received limited attention. Thus, focusing on natural disasters as an environmental factor that alters the level of managerial risk taking, we theorize how such disasters affect firm search orientation. Specifically, we argue that experiencing a natural disaster leads firm decision makers to engage in less managerial risk taking by increasing their perceived variance of outcomes (particularly, the likelihood of maximum loss), reducing their confidence in their capabilities to foresee the future, and arousing anxiety and stress. As a result, this diminished risk preference directs search activities more towards exploitation rather than exploration, because the returns from exploration are often more uncertain and distant compared to those from exploitation. To validate our hypothesis, we utilize the Great East Japan Earthquake as a natural experiment, an event recognized as the deadliest and most devastating in Japanese history.
The Voice Experience Inventory (VOXI): Validating a consensus-driven instrument for measuring user impressions of computer voice
No full textVoice interaction has entered daily life. Yet, the measurement of voice user experience (voice UX) remains underdeveloped. We took a consensus-based approach to the creation of a new instrument grounded in the work of voice UX researchers. For this, we conducted a systemic review of the literature from ACM, IEEE, and Web of Science (N = 49) to gather a corpus of instruments, measures, and methodological perspectives. We then carried out two validation studies on a range of computer voices, showing through exploratory (N = 100) and confirmatory factor analyses (N = 150) that five underlying factors unify the consensus-derived measures. We offer the 24-item Voice Experience Inventory (VOXI) alongside three comprehensive frameworks on key independent and dependent variables and their relationships, anchored to the material that guided our instrument development process. This work represents a timely and community-centred path towards rigorous and standardized measurement of voice UX.
Bioinspired trajectory modulation for effective slip control in robot manipulation
No full textEnsuring a stable grasp during robotic manipulation is essential for dexterous and reliable performance. Traditionally, slip control has relied on grip force modulation. Here we show that trajectory modulation provides an effective alternative for slip prevention in certain robotic manipulation tasks. We develop and compare a slip control policy based on trajectory modulation with a conventional grip-force-based approach. Our results demonstrate that trajectory modulation can significantly outperform grip force control in specific scenarios, highlighting its potential as a robust slip control strategy. Furthermore, we show that, similar to humans, incorporating a data-driven action-conditioned forward model within a model predictive control framework is key to optimizing trajectory modulation for slip prevention. These findings introduce a predictive control framework leveraging trajectory adaptation, offering a new perspective on slip mitigation. This approach enhances grasp stability in dynamic and unstructured environments, improving the adaptability of robotic systems across various applications.
광 중합체 합성
No full textThe present disclosure relates to a photopolymer composition, and more particularly, to a compound having a novel structure, a photopolymer composition including the compound as a dye, a hologram recording medium produced from the photopolymer composition, an optical element including the hologram recording medium, and a holographic recording method using the photopolymer composition
Shoulder-assisted wearable robot system applied with artificial muscles and artificial joints
No full text상지 착용형 로봇으로서, 착용자의 상지 움직임에 따라 수축되거나 이완되어 힘을 생성하는 인공 근육, 착용자의 어깨뼈에 상기 상지 착용형 로봇을 고정시키는 어깨 착용부, 착용자의 상완에 위치하며, 인공 근육에서 생성된 힘을 힘 전달부 또는 인공 관절부로 전달하는 상완 착용부, 인공 근육과 상완 착용부 사이를 연결하며, 인공 근육에서 발생되는 힘을 착용자의 상지에 전달하는 힘 전달부, 그리고 어깨 착용부와 상완 착용부를 연결하며 착용자의 어깨 외전 각도를 기초로 인공 근육에서 발생하는 힘과 상완 사이의 각도를 조절하는 인공 관절부로 구현된다