163559 research outputs found
Sort by
Can Citizens Avoid the Legacy News Media and Still be Informed?
No full text최근 기성언론에 대한 신뢰 하락을 둘러싸고 우려가 높지만 이 같은 변화가 민주주의 구성원으로서 시민들의 삶에 어떤 영향을 미칠지에 대해 실증적으로 살펴본 시도는 많지 않았다. 본 연구는 기성언론에 대한 불신이 시민들의 정치지식 수준에 어떤 영향을 미칠지 살펴봄으로써 식견있는 시민(informed citizen) 양성이라는 언론의 전통적인 규범 수행에 어떤 변화가 있을지 알아보고자 했다. 연구자들은 기성언론에 대한 신뢰 하락에 따라 시민들이 기성언론은 선호하지 않는 동시에 대안적 뉴스채널에 의존하는 뉴스 이용방식을 기성언론회피형 뉴스소비성향이라고 개념화하고 이 성향이 언론 신뢰 하락과 정치지식 간의 관계를 설명하는 핵심적 요인이라고 이해했다. 구체적으로 연구자들은 1) 기성언론에 대한 불신이 기성언론회피형 소셜미디어 뉴스소비와 어떤 관계가 있는지, 2) 기성언론회피형 소셜미디어 뉴스소비성향은 기성언론 불신과 정치지식의 관계를 어떻게 매개하는지 살펴봤다. 연구 결과 기성언론에 대한 불신이 높은 시민들은 기성언론회피형 소셜미디어 뉴스소비성향이 높았다. 기성언론은 회피하고 소셜미디어에 의존해 정보를 습득하는 시민들은 특별히 객관적 정치지식 수준이 높지 않았지만 스스로 정치에 대해 잘 알고 있다고 여기는 정도를 의미하는 주관적 정치지식 수준이 높았다. 기성언론회피형 소셜미디어 뉴스소비성향은 정치지식과신 정도와 정적인 관계에 있었고 기성언론에 대한 불신과 주관적 정치지식 및 정치지식과신의 관계를 정적으로 매개했다.
Researchers seek to explore what changes would be made to the news medias traditional norm of cultivating informed citizen by analyzing how distrust in the legacy news media would be associated with the level of political knowledge among citizens. Researchers identified legacy media-avoidant news consumption as a key factor that could explain the relationship between the decline in trust in news media and political knowledge. Specifically, the researchers examined 1) how distrust in legacy media is associated with legacy media-avoidant news consumption, and 2) how legacy media-avoidant news consumption mediates the relationship between distrust in legacy news media and political knowledge. Findings show that citizens with a high level of distrust in the legacy media are more likely to consume news from alternative sources. While citizens who avoid legacy media and rely on social media did not demonstrate a particularly high level of objective political knowledge, they did demonstrate a high level of subjective political knowledge and overconfidence in political knowledge. The tendency to avoid legacy news media when consuming news positively mediates the relationship between distrust in legacy news media and subjective political knowledge as well as overconfidence in political knowledge
Stakeholders Perceptions of Journalism Activities and Regulation about Cable TV Local Channels: A Case Study of KCTV Jeju Broadcasting
No full text종합유선방송사업자(SO)이자 지역채널인 KCTV제주방송은 제주도민과 시청자, 그리고 방송연구자들 사이에서 저널리즘 역할을 잘 수행하는 지역방송국으로 회자되곤 한다. 하지만 이를 뒷받침할 관련 연구나 지표는 부족한 상황이다. 이에 본 연구는 KCTV 내부종사자와 외부관계자인 지상파 방송 종사자, 지역언론 연구자를 대상으로 심층인터뷰를 진행하여 지역방송국으로서 KCTV의 저널리즘 행위에 대해 살펴보았다. 연구결과는 다음과 같다. 우선 타 방송사와 달리 중앙방송에 종속되지 않는, 즉 태생적 특성에 기반한 편성의 자율성과 충분한 방송시간 확보가 지역성 구현의 주요 요인으로 인식되고 있었다. 한편, 케이블 채널에만 부과된 해설·논평 금지와 같은 규제에 대해 내부종사자들은 구시대적 발상이자 폐지되어야 할 사문화된 조항이라 여기고 있었다. 외부관계자들은 케이블TV의 저널리즘 행위가 공공성 구현에 오히려 부정적인 영향을 줄 수 있다고 보면서도, KCTV의 24시간 자체 편성과 광케이블을 활용한 실시간 방송을 지역성 구현을 최적화 할 수 있는 요인으로 받아들이고 있었다. 종합적으로 KCTV가 해설·논평 금지라는 규제와 상대적 공적 지원의 미비에도 불구하고 편성의 자율권과 다양한 시청자 접근권 확대 등 지역민 친화적인 하이퍼로컬 미디어로서의 특성을 보여주고 있음을 확인할 수 있었다.
As a comprehensive cable television operator (SO) and local channel,
KCTV Jeju Broadcasting is often regarded among Jeju residents,
viewers, and broadcasting researchers as a local broadcaster that
performs its journalistic role well. However, there is a lack of related
research or indicators to support this reputation. Therefore, this study
conducted in-depth interviews with internal KCTV employees and
external stakeholders, including terrestrial broadcasting employees
and local media researchers, to examine KCTV's journalistic activities
as a local broadcaster. The results of the study are as follows. First,
unlike other broadcasters, the autonomy of programming based on
its inherent characteristics and the securing of sufficient broadcasting
time were recognized as major factors in realizing locality. Meanwhile,
internal employees considered regulations imposed only on cable
channels, such as the 'ban on commentary and analysis,' as 'outdated
concepts' and 'dead clauses' that should be abolished. External
stakeholders acknowledged that the journalistic activities of cable TV
could negatively impact the realization of public interest, but they
also accepted that KCTV's 24-hour self-programming and real-time
broadcasting utilizing fiber optic cables could optimize the realization
of locality. Overall, it was confirmed that KCTV, despite regulations
such as the ban on commentary and analysis and the relative lack of
public support, demonstrates characteristics of a hyperlocal media
that is friendly to local residents, such as programming autonomy and
expanding various viewer access rights
Recent Advances in Hydrogel-Based Soft Bioelectronics and its Convergence with Machine Learning
No full textRecent advancements in artificial intelligence (AI) technologies, particularly machine learning (ML) techniques, have opened up a promising frontier in the development of intelligent soft bioelectronics, demonstrating unparalleled performance in interfacing with the human body. Hydrogels, owing to their unique combination of biocompatibility, tunable mechanical properties, and high water content, have emerged as a versatile platform for constructing soft bioelectronic devices. Functionalized hydrogels, such as conductive hydrogels, can efficiently capture biosignals from various target tissues while seamlessly forming conformal and reliable interfaces. They can also function as an intermediary layer between biological tissues and soft bioelectronics for diagnosis and therapy purposes. Meanwhile, ML has demonstrated its efficacy in processing extensive datasets collected from the bioelectronics. The convergence of hydrogel-based soft bioelectronics and ML has unlocked a myriad of possibilities in unprecedented diagnostics, therapeutics, and beyond. In this review, the latest advances in hydrogel-based soft bioelectronics are introduced. After briefly describing the materials and device strategies for high-performance hydrogel bioelectronics, how ML can be integrated to augment the functionalities is discussed. Recent examples of ML-integrated hydrogel bioelectronics are then discussed. Finally, the review is concluded by introducing future potential applications of AI in hydrogel-based bioelectronics, alongside inherent challenges in this interdisciplinary domain. The integration of machine learning techniques with hydrogel-based soft bioelectronics is opening a new avenue for intelligent biomedical devices and systems. The hydrogel-based bioelectronics can achieve seamless integration with human biological tissues, thus achieving reliable and sustainable operation. This facilitates high-quality biosignal acquisition, which, in turn, is used for training the artificial intelligence for efficient data processing and system design.image (c) 2024 WILEY-VCH GmbHN
Soft Implantable Bioelectronics for the Management of Neurological Disorders and Cardiovascular Diseases
No full textContinuous monitoring systems that effectively manage neurological disorders and cardiovascular diseases are crucial for enhancing the quality of life in aging societies. Soft implantable bioelectronics, which consists of soft devices with mechanical properties akin to biological tissues, are promising candidates for such monitoring systems. These devices enable conformal integration with target tissues, minimizing adverse side effects during their long-term use. They also precisely monitor physiological conditions, facilitating tailored therapeutic interventions based on individual physiological responses. This review highlights recent advancements in soft implantable bioelectronics for managing neurological disorders and cardiovascular diseases. Specifically, strategies for the materials, device fabrication, and their integration with the biological tissues are reviewed in detail first. Then, the challenges in translating these technologies into practical applications are addressed, including the integration of artificial intelligence (AI) technologies.N
행동 및 피질에서 양가적 경험으로 특정되는 경외감
No full text학위논문(석사) -- 서울대학교 대학원 : 자연과학대학 뇌인지과학과, 2024. 8. 차지욱.The ambivalent nature has been emphasized as a unique quality of awe in qualitative descriptions and cited as a potential source of its various psychiatric and psychosocial benefits. However, due to the affective sciences dichotomic positive/negative schema, this ambivalence has not been fully explored. This study aims to capture the valence dynamics inherent in awe by applying naturalistic VR-EEG paradigm and an extended valence measurement allowing ambivalent response. We investigated whether awe can be better characterized as ambivalent rather than simply positive or negative at both behavioral and cortical levels. Behaviorally, the awe intensity for each clip was precisely predicted by the duration and intensity of ambivalent feelings, but not by other valence metrics. In the cortical level, ambivalent feelings during awe showed unique neural representations in the latent cortical space, with significant individual variability in their distinctiveness from positive/negative representations. Nevertheless, the more distinctly ambivalent feelings were encoded in the cortex, the stronger individuals reported awe. Finally, frontal delta band power was mainly involved in distinguishing different valence representations in the cortices. This study not only explores the existence of unique neural representations of ambivalent feelings, a topic of debate in affective neuroscience, but also demonstrates that awe can be characterized as an ambivalent experience at both behavioral and cortical levels.경외감에 동반하는 양가적 느낌은 경외 체험에 대한 질적 묘사에서 독자적인 성질로 강조되어 왔으며, 그것이 갖는 다양한 정신의학적 및 심리사회적 이점의 가능한 근원으로 언급되어 왔다. 그럼에도 불구하고 정서과학의 이분법적인 긍/부정 도식으로 인해 경외감의 양가적 성질은 온전히 연구되지 못했다. 본 연구에서는 경외에 내포된 유인성의 동역학을 풍부하게 포착하기 위한 가상현실-뇌전도 결합 프로토콜과 양가적 느낌의 응답을 허용하는 확장된 유인성 측정도구를 사용하여, 행동 및 피질 수준에서 경외감이 단순 긍/부정 느낌보다 양가적 느낌으로 더 정확히 특정될 수 있는지 살펴보았다. 행동 수준에서, 참여자들이 각 가상현실 영상에 대해 평정한 경외 수준은 양가적 느낌의 길이와 강도에 의해 정확히 예측될 수 있었던 반면, 다른 유인성에 관한 평정값으로는 유의하게 예측되지 못했다. 피질 수준에서, 경외 체험 동안의 양가적 느낌은 잠재 신경 공간에서 독자적인 신경 표상을 보였지만, 그 신경 표상이 긍/부정 느낌의 표상에 대해 갖는 구분가능성은 큰 개인차를 보였다. 그럼에도 불구하고, 양가적 느낌이 긍/부정 느낌의 표상과 질적으로 구분되어 피질에 부호화 될수록, 더 강한 경외감이 보고되었다. 마지막으로, 다른 종류의 유인성 표상을 구분하는데 있어 주로 전측 영역의 델타파의 파워가 관여하는 것으로 관찰되었다. 이 연구는 정서 신경과학에서 논쟁되어오던 양가적 느낌의 독자적인 신경표상을 탐색했을 뿐 아니라, 이를 기반으로 행동 및 피질 수준에서 경외감이 양가적 경험으로써 더욱 정확히 특정될 수 있음을 밝혔다는 점에서 의의를 갖는다.Abstract i
Introduction 1
Methods
Participants 6
Experimental paradigms 6
Behavioral analysis 13
Electrophysiological analysis 14
Results
Awe experience is consistently associated with longer and stronger ambivalent feelings, but not with single-valence feelings 23
Ambivalent feelings predict the awe intensity more precisely than other single valence metrics 26
Aligned latent cortical spaces share valence representation architecture across individuals and stimuli 26
The more distinctively ambivalent feelings are represented in the cortices, the more saliently individuals experience awe 30
The delta oscillation in the frontal channels mainly engages in distinguishing different valence representation 34
Discussion 36
References 43
Supplementary Materials 52
Acknowledgement 54
Abstract in Korean 55석
How Gift-Giving Affects Consumer Indulgence
No full text학위논문(석사) -- 서울대학교 대학원 : 경영대학 경영학과, 2024. 8. Youjae Yi.This study explores the diverse effects of gift-giving on indulgent consumption choices based on different giving motivations. Drawing on the Hedonic Contingency Model and mood maintenance theory, we examine how altruistic motives behind gift-giving elicit positive emotions, potentially driving givers toward indulgent consumption to sustain these emotional highs. Additionally, we investigate the role of perceived pain of payment in gift-giving, grounded in loss aversion theory. Obligatory gift-giving may trigger self-control mechanisms, leading givers to prefer less indulgent consumption behaviors to mitigate perceived losses. We also explore how recipient responses, particularly gratitude, influence giver behavior by mitigating the negative impacts of obligatory giving. Through this research, we aim to illuminate the complex dynamics between gift-giving and consumer behavior, providing valuable insights into how acts of kindness shape consumption preferences. This study contributes to a deeper understanding of the psychological mechanisms underlying gift- giving and its implications for consumer decision-making.본 연구는 다양한 동기에 따른 선물 행위가 탐닉적 소비 선택에 미치는 영향을 탐구한다. 쾌락적 우발 모델 (Hedonic Contingency Model)과 기분 유지 이론(Mood Maintenance Theory)에 의거하여, 이타적 동기로 인한 선물 행위가 어떻게 선물 행위자에게 긍정적인 감정을 유발하는지, 그리고 잠재적으로 선물 행위자가 이러한 감정적 고양감을 유지하기 위해 탐닉적 소비로 유도되는지 살펴본다. 또한, 손실 회피 이론(Loss Aversion Theory)에 근거하여 선물 행위에서 지각된 지불의 고통의 역할을 조사한다. 의무적인 선물 행위는 자기 통제 메커니즘을 촉발시켜 선물 행위자가 지각된 손실을 줄이기 위해 덜 탐닉적인 소비 행동을 선호하게 만들 수 있다. 덧붙여, 의무적인 선물 행위의 부정적 영향을 완화하는 특히, 감사와 같은 수혜자의 반응이 어떻게 선물 행위자의 행동에 영향을 주는지 탐구한다. 본 연구를 통해 선물 행위와 소비자 행동 간의 복잡한 역학을 밝히고자 하며, 친절한 행위가 소비자 행동에 미치는 영향을 이해하는 데 귀중한 통찰력을 제공하고자 한다. 본 연구는 선물 행위에 내재된 심리적 메커니즘과 그것이 소비자 의사 결정에 미치는 영향을 더 깊이 이해하는 데 기여한다.Chapter 1. Introduction 1
Chapter 2. Literature Review and Hypotheses 3
2.1 Gift-Giving and Positive Feelings
2.2 Goal to Maintain Positive Feelings and Self-Indulgence
2.3 Motivations of Giving
2.4 Negative Effect of Gift-Giving on Indulgence
2.5 Perceived Pain of Payment, Less Indulgence
2.6 Perceived Gratitude
Chapter 3. Study 1 13
3.1 Participants and Procedure
3.2 Dependent Measures
3.3 Results
3.4 Discussion
Chapter 4. Study 2 16
4.1 Participants and Procedure
4.2 Dependent Measures
4.3 Results
4.4 Discussion
Chapter 5. General Discussion 21
5.1 Conclusion
5.2 Managerial Implication
5.3 Limitation and Future Research
Bibliography 26
Abstract in Korean 34석
Dkk2 과발현이 마우스 행동과 신경기능에 미치는 영향
No full text학위논문(석사) -- 서울대학교 대학원 : 의과대학 의과학과, 2024. 8. 김명환.The hippocampus regulates cognitive function, anxiety, and social behavior, while its dysfunction has been implicated in psychiatric disorders such as depression, addiction, and schizophrenia. Dickkopf-related protein 2 (Dkk2) is known as a modulator of Wnt signaling, which plays a critical role in the development of the nervous system and is associated with psychiatric disorders. It has been suggested that Dkk2 bidirectionally regulates Wnt signaling in an organ-specific manner. However, the role of Dkk2 in the nervous system remains unknown. In this study, I investigated whether altered Dkk2 expression levels influence molecular and cellular signaling in the hippocampus and mouse behavior. Using transgenic (TG) mice overexpressing Dkk2 in neurons, I found that increased Dkk2 expression decreased CaMKII phosphorylation in the hippocampus. Behavioral analyses revealed that Dkk2 transgenic mice exhibited anxiolytic-like behaviors and impaired social recognition. These behavioral changes were accompanied by increased neurogenesis in the hippocampal dentate gyrus (DG). Dkk2 TG mice showed an increased density of the Dcx-positive cells in the DG subfield compared to WT mice. However, increased Dkk2 levels did not affect the intrinsic excitability of DG granule cells. Resting membrane potential, spike firing, and input resistance did not differ between WT and TG mice. Overall, these results suggest that Dkk2 acts as a negative regulator of Wnt signaling in the nervous system, and increased Dkk2 levels in neurons enhance neurogenesis in the DG and affect animal behavior.해마는 인지 기능, 불안 및 사회적 행동을 조절하며, 해마의 기능이상은 우울증, 중독 및 조현병과 같은 정신 질환과 관련되어 있다고 알려져 있다. 딕코프 관련 단백질 2 (Dkk2)는 정신질환과 관련성이 알려진 Wnt 신호 전달의 조절인자로 알려져 있고, 장기에 따라 Wnt 신호를 항진 또는 억제하는 것으로 알려져 있다. 그러나 중추신경계에서는 Dkk2가 Wnt 신호를 어떻게 조절하는지 알려지지 않았다. 본 연구에서는 Dkk2 과발현에 따른 해마의 분자 및 세포 수준의 변화와 마우스의 행동적 변화에 대해 연구를 하였다. 증가된 Dkk2 발현은 해마에서 CaMKII 인산화를 감소함을 발견하였다. 행동 실험을 통해 Dkk2 과발현 마우스에서 불안 행동의 감소와 사회적 인식 장애를 관찰하였다. 증가된 Dkk2 발현이 성체신경생성에 어떤 영향을 미치는지 조사하기 위해 치상회 (dentate gyrus, DG) 과립세포에서 Dcx 양성 세포 밀도를 비교하였다. Dkk2 형질전환 마우스는 WT 마우스에 비해 치상회에서 성체신경생성이 증가하였다. 그러나, Dkk2의 발현 증가는 치상과립세포 (DG granule cells)의 흥분성, 휴지막 전위, 및 입력 저항에는 영향이 없었다. 상기한 결과들은 뉴런 특이적으로 과발현된 Dkk2가 칼슘/Wnt 경로를 억제하고 성체신경세포생성을 증가시키며, 해마와 관련된 이상 행동을 유발할 수 있음을 시사한다.Abstract - 1 -
Contents · - 3 -
List of Figures -４-
Introduction -６-
Methods -10-
Results -18-
Discussion · -41-
References -44-
국문초록 -49-석
나노표면구조기반 소형 수분포텐셜센서 개발
No full text학위논문(박사) -- 서울대학교 대학원 : 공과대학 기계공학부, 2024. 8. 고승환.Soil moisture plays an indispensable role in the dynamics of energy and mass transfer within SPAC. The characteristics of soil moisture are defined by the existing water content (Soil Water Content) and its capacity to retain moisture (Soil Water Potential). The movement of moisture transpires from areas with higher soil water potential to those with lower soil water potential, ultimately reaching a state of equilibrium with its surrounding environment. This flux is governed by variations in water potential energy across different locations. While soil water content provides insight into the volume of water present, soil water potential elucidates the water's availability for uptake by plants, microbial activities, and its transport through the soil matrix, thereby underscoring the soil's ability to retain and release water within the SPAC framework. Consequently, soil water potential emerges as a critical parameter in soil sciences that necessitates rigorous quantitative analysis. Over the past thirty years, considerable advancements have been achieved in techniques and mathematical models to effectively measure and predict soil water potential. For instance, the advent of partial differential equations has significantly contributed to modeling water movement in unsaturated soils, reformulated into diffusion-type equations to elucidate three-dimensional unsaturated flow dynamics. Recent advancements have integrated soil water potential into the SPAC conceptual framework, providing insights into how plants physiologically, morphologically, phytochemically, and phytopathologically react to variations in soil moisture. However, research on soil water potential remains less developed compared to soil water content, largely hindered by the absence of effective and user-friendly measurement methodologies. This study introduces a micro water potential sensor system specifically designed to accurately and efficiently measure the full spectrum of soil water potential required for optimal plant growth. The sensor system consists of multiple modules, including a sensor component for real-time monitoring of soil water potential fluctuations in actual field conditions, leveraging Printed Circuit Board (PCB) and micro/nano fabrication technologies. Additionally, it incorporates a capacitive measurement component for dew detection and a high-precision temperature measuring subsection for dew point determination. Calibration of the individual sensors was conducted utilizing a standard water potential solution (NaCl), resulting in consistent signal distributions across the sensors. The performance of the sensor was evaluated within the soil moisture potential range of -0.033 MPa to -1.5 MPa, the primary range from which plants absorb water. Successful measurements were recorded during soil testing within this range. Additionally, continuous data collection was carried out throughout the drying process by deploying both the soil water potential sensor and the soil moisture sensor in the same soil over several days. A soil water characteristic curve was constructed, demonstrating a rise in the magnitude of soil water potential as soil moisture content decreased. This continuous monitoring system for water potential is expected to significantly enhance water management practices in agriculture by enabling precise tracking of plant water stress in the soil.ABSTRACT I
TABLE OF CONTENTS IV
LIST OF FIGURES VI
CHAPTER 1. GENERAL INTRODUCTION １
1.1. WATER POTENTIAL １
1.1.1. Significance of Water Potential Measurement for SPAC １
1.1.2. Comprehending the elements of soil water potential ４
1.1.3. Traditional methods for measuring soil water potential ７
1.1.4. Simulation methods for measuring soil water potential １０
1.2. RESEARCH OBJECTIVE １２
BIBLIOGRAPHY １４
CHAPTER 2. METHODS: SENSOR DESIGN, FABRICATION AND MEASUREMENT SET UP １８
2.1. DESIGN OF MICRO WATER POTENTIAL SENSOR １８
2.1.1. Structure design of the water potential sensor １８
2.1.2. Fabrication process of water potential sensor １９
2.1.3. Nano-rod patterning process ２１
2.1.4. Fabrication result ２４
2.2. MEASUREMENT SETUP ２８
2.2.1. Water potential measurement module ２８
BIBLIOGRAPHY ３１
CHAPTER 3. THEORY AND ANALYSIS ３２
3.1. THE SOIL WATER CHARACTERISTIC ３２
3.1.1. The relationship between soil water potential and water content ３２
3.2. DEW POINT WATER POTENTIAL SENSOR ３９
3.2.1. Theoretical background and assumptions ３９
3.2.2. Measurement Principle ４２
3.3. WATER MOLECULE ADSORPTION CHARACTERISTICS ４４
OF NANO TURF ４４
3.3.1. Chemisorption and physisorption of water on the TiO2 ４４
BIBLIOGRAPHY ４８
CHAPTER 4. SENSOR CALIBRATION AND THERMAL CHARACTERISTIC EVALUATION ５２
4.1. SENSOR CALIBRATION ５２
4.1.1. Water potential sensor calibration ５２
4.2. WATER POTENTIAL SENSOR ANALYSIS ５４
4.2.1. Sensor response test and characteristics ５４
4.2.2. Temperature effect ５６
BIBLIOGRAPHY ５８
CHAPTER 5. FIELD MEASUREMENT OF SOIL WATER POTENTIAL SENSOR ６３
5.1. FIELD MEASUREMENT OF THE WATER POTENTIAL SENSOR ６３
5.1.1. Installation set up ６３
5.1.2. Water potential signal from a drying soil ６４
5.1.3. Comparison result with water activity measurement system ６５
BIBLIOGRAPHY ６９
CHAPTER 6. CONCLUSIONS ７０박
부분적으로 관측된 특징을 가진 그래프를 위한 확산 기반 특징 대치법
No full text학위논문(박사) -- 서울대학교 대학원 : 공과대학 전기·정보공학부, 2024. 8. 최진영.In recent years, graph neural networks (GNNs) have made tremendous progress in graph learning tasks. However, GNNs typically assume that features for all nodes are fully observed, an assumption that does not hold in many real-world situations. For various reasons, graph-structured data often contains missing features, which hinders the performance of GNNs that exhibit powerful performance in graph learning tasks. To address this issue, this thesis proposes four feature imputation schemes. Specifically, all the proposed schemes diffuse known feature values along edges to impute missing features. These diffusion-based imputation schemes demonstrate superiority across diverse data structure, ranging from graph-structured data to matrix-formatted scRNA-seq data. First, a scheme called pseudo-confidence-based feature imputation defines pseudo-confidence and leverages dependency between channels in its imputation process. Second, we identify a common problem that all diffusion-based methods suffer from and improve performance in downstream tasks by addressing this issue. Third, a new scheme called heterogeneous graph feature imputation is the first attempt to utilize diffusion-based imputation on heterogeneous graphs. Lastly, we apply diffusion-based imputation to scRNA-seq data by designing a method based on genetic evidence. Overall, the research presented in this thesis facilitates the use of off-the-shelf GNNs in real-world scenarios with missing features, preventing performance degradation in downstream tasks.최근 몇 년간 그래프 신경망(GNN)은 그래프 학습 태스크에서 엄청난 진전을 이루었다. 그러나 GNN은 일반적으로 전체 노드의 특징이 완전히 관찰된 것으로 가정하는데, 이러한 가정은 많은 현실 세계 상황에서 성립하지 않는다. 다양한 이유로 그래프 구조의 데이터는 유실된 특징을 포함하고, 이는 그래프 학습 태스크에서 강력한 성능을 보이는 GNN의 이용을 방해한다. 이 문제를 해결하기 위해 본 논문에서는 세 가지 특징 대치법을 제안한다. 구체적으로, 제안된 모든 방법은 알려진 특징 값들을 엣지를 따라 확산시켜 유실된 특징을 대치한다. 이러한 확산 기반 대체 방식은 그래프 구조 데이터에서 행렬 형태의 scRNA-seq 데이터에 이르기까지 다양한 데이터 구조에서 우월성을 보인다. 첫째로, 수도 신뢰도 기반 특징 대치라는 방법은 수도 신뢰도를 정의하고, 이 대치 과정에서 채널 간 의존성을 활용한다. 둘째로, 모든 확산 기반 방법이 겪는 문제점을 발견하고, 이 문제를 해결함으로써 하위 태스크에서의 성능을 개선한다. 셋째로, 이질적 그래프 특징 대치라는 방법은 이질적 그래프에서 확산 기반 대치를 처음으로 시도한다. 마지막으로, 유전적 근거에 기반한 방법을 설계하여 scRNA-seq 데이터에 확산 기반 대치를 적용한다. 정리하자면, 본 논문에서 제시된 연구는 유실된 특징이 있는 실제 상황에서도 개발된 GNN의 활용을 가능하게 하여 하위 태스크에서의 성능 저하를 방지한다.Abstract i
Contents ii
List of Tables vi
List of Figures ix
1 INTRODUCTION 1
1.1 Motivation and Background 1
1.2 Problem Statement 2
1.3 Contributions 2
2 RELATED WORK 4
2.1 Learning on Graphs with Missing Features 4
2.1.1 GNN Architecture 4
2.1.2 Reconstruction 4
2.1.3 Feature Propagation 5
2.2 Manifold Learning 5
2.3 Graph Diffusion 6
2.4 Distance Encoding 6
2.5 Representation Learning on Heterogeneous Graphs 7
2.6 Handling noise in scRNA-seq data 7
2.7 Propagation-based imputation in scRNA-seq data 8
3 Pseudo-Confidence-based Feature Imputation for Graphs with Partially Known Features 10
3.1 Overview 10
3.2 Pseudo-Confidence-based Feature Imputation 12
3.2.1 Notations 13
3.2.2 PC-based Feature Imputation 14
3.2.3 Pseudo-Confidence 15
3.2.4 Channel-Wise Inter-Node Diffusion 16
3.2.5 Node-Wise Inter-Channel Propagation 19
3.3 Experiments 21
3.3.1 Experimental Setup 21
3.3.2 Semi-Supervised Node Classification Results 24
3.3.3 Link Prediction Results 25
3.3.4 Classification Accuracy According to Pseudo-Confidence 25
3.3.5 Degree of Feature Recovery According to Pseudo-Confidence 26
3.4 Proof 27
3.4.1 Proof of Proposition 3.2.1 27
3.4.2 Convergence of Channel-wise inter-node Diffusion 28
3.5 Summary 31
4 Diffusion with Synthetic Features: Feature Imputation for Graphs with Partially Observed Features 32
4.1 Overview 32
4.2 Notation and Problem Definition 34
4.3 Feature Imputation with Synthetic Features 36
4.3.1 Overview of FISF 36
4.3.2 Pre-diffusion 37
4.3.3 Synthetic Feature Generation 38
4.3.4 Diffusion with Synthetic Features 39
4.4 Experiments 42
4.4.1 Datasets and Baselines 42
4.4.2 Experimental Setup 43
4.4.3 Semi-supervised Node Classification Results 45
4.4.4 Link Prediction Results 46
4.4.5 Ablation Study 47
4.4.6 Missing Not at Random Setting 48
4.4.7 Complexity Analysis 48
4.4.8 Scalability of FISF 50
4.4.9 Contribution of Low-variance Channels in Downstream Tasks 51
4.4.10 Effects of Hyperparameters 53
4.4.11 Hyperparameter search for FISF 53
4.4.12 Smoothness Analysis 55
4.4.13 Synthetic Features Sampled from a Non-Uniform Distribution 57
4.4.14 Zero Initialization vs Random Initialization 59
4.5 Proof of Convergence of Diffusion Stages 59
4.6 Summary 62
5 Edge-Type-Aware Diffusion for Heterogeneous Graphs with Partially Observed Features 63
5.1 Overview 63
5.2 Problem Setting 66
5.3 Heterogeneous Graph Feature Diffusion 68
5.3.1 Overview of HetGFD 68
5.3.2 Preliminary Diffusion 68
5.3.3 Edge-Type-Wise Homophily 70
5.3.4 Edge-Type-Aware Diffusion 72
5.4 Experiments 74
5.4.1 Experimental Setup 74
5.4.2 Semi-Supervised Node Classification Results 77
5.4.3 Link Prediction Results 79
5.4.4 Ablation Study 79
5.5 Proof 81
5.6 Summary 83
6 Gene-Gene Relationship Modeling Based on Genetic Evidence for Single-Cell RNA-Seq Data Imputation 85
6.1 Overview 85
6.2 Related Work 87
6.3 Preliminaries 87
6.4 Proof of Convergence of Feature Propagation 91
6.5 Proposed Method 94
6.5.1 Overview of scCR 94
6.5.2 Pre-imputation Stage 94
6.5.3 Complete Relationship Stage 95
6.5.4 Denoising Stage 98
6.6 Experiments 99
6.6.1 Experimental Setup 99
6.6.2 Results 102
6.7 Summary 105
7 CONCLUSION 107
Abstract (In Korean) 124
v박
A Feasibility Study
No full text학위논문(석사) -- 서울대학교 대학원 : 사범대학 체육교육과, 2024. 8. Song Wook.Purpose: Breast cancer rates are increasing, particularly among young generation. With advancements in medical technology and treatments, the survival rates among breast cancer patients are steadily increasing. While exercise has proven benefits for menopausal symptoms, physical function, and quality of life in breast cancer survivors, there is limited research on its impact for young survivors. The primary purpose of this study is to assess the feasibility and usability of Virtual Mate Me (VMM) app and the secondary purpose of this study is to evaluate the impact of the exercise program in improving early menopause symptoms, physical function, and quality of life on young breast cancer survivors. Methods: This feasibility study involved pre-menopausal female survivors (n = 30) who were recruited and randomly assigned to either the exercise group or the control group. The exercise group participated in 24 sessions of exercise through the VMM app meanwhile the control group was instructed to maintain their lifestyle routine as usual for 8 weeks. The primary outcome was to assess the feasibility of the VMM app by tracking the retention rates (at least ³80%), adherence rates (at least ³75%), and usability rates. Secondary outcomes included early menopause symptoms, physical functions (cardiorespiratory fitness, isokinetic knee strength and endurance, handgrip strength, shoulder range of motion, upper limb disability questionnaire), level of physical activity, and quality of life. Results: 22 out of 30 participants, aged 38.59 ± 5.89 (mean ± SD) years old completed this study. The retention rate (73.33%) did not meet the target (at least 80%), but adherence rate (83.32%) achieved the target of at least 75%, or 18/22 sessions. The usability of the VMM app resulted in above-average ratings. For exercise-related effects, upper limb disabilities (DASH score) significantly decreased after the exercise intervention (p = 0.013) in EXG, and 180o flexion in isokinetic strength significantly improved in left leg (p = 0.037). Conclusion: This study suggests that the VMM apps is feasible and usable for future use. Additionally, this exercise program could alleviate upper limb disabilities for breast cancer survivors underwent surgery and improve lower limb strength.
Keyword : Young breast cancer, Exercise, Mobile-apps, Digital healthcare, Menopause symptoms Student Number : 2022-29148목적: 유방암 발병률은 특히 젊은 세대에서 증가하고 있다.의료 기술과 치료법의 발전으로 유방암 환자의 생존율은 지속적으로 증가하고 있다. 운동이 폐경 증상, 신체 기능 및 삶의 질에 미치는 이점이 입증되었지만,젊은 생존자에게 미치는 영향에 대한 연구는 제한적이다.본 연구의 주 목적은 Virtual MateMe (VMM)앱의 실현 가능성과 유용성을 평가하는 것이며,부차적 목적은 VMM앱을 통해 수행한 운동 프로그램이 젊은 유방암 생존자의 조기 폐경 증상,신체 기능 및 삶의 질 개선에 미치는 영향을 평가하는 것이다.
방법: 본 연구에는 30 명의 폐경 전 여성 유방암 생존자가 참여하였으며,운동 그룹(EXG)또는 통제 그룹 중 하나로 무작위 배정되었다.운동 그룹은 VMM앱을 통해 8 주간 총 24 회의 운동 세션에 참여하였으며, 통제 그룹은 8 주간 평소 생활양식을 유지하도록 하였다.일차적 변수로는 VMM앱의 실현 가능성을 추적하고자 연구 참여 지속률 (80%이상),운동 참여율 (75%이상)및 유용성을 평가하였다.이차적 변수에는 조기 폐경 증상,신체 기능(심폐지구력, 등속성 무릎관절 근력 및 근지구력,악력,어깨 가동 범위,상지 기능장애 설문지),신체 활동 수준 및 삶의 질을 포함하였다.
결과: 30 명의 참가자 중 22 명이 본 연구를 완료하였으며,평균 연령은 38.59± 5.89 세였다. 연구 참여 지속률(73.33%)은 목표치(80%)에 미치지 못하였지만, 운동 참여율(83.32%)은 목표치인 75%(24 회 중 18 회)이상을 달성하였다. VMM앱의 유용성은 평균 이상의 등급을 받았다. 운동 프로그램 중재 후,상지 기능장애 설문지 점수가 EXG 에서 유의하게 감소하였으며, 등속성 무릎관절 근지구력 측정에서 왼쪽 다리의 굴곡력이 유의하게 향상되었다(p=0.013,p =.037).
결론: 본 연구의 결과는 추후 VMM앱이 실현 가능하며,젊은 유방암 생존자에게 유용할 것임을 시사한다. 또한,이 운동 프로그램은 수술을 받은 유방암 생존자의 상지 기능장애를 완화하고,하지의 근기능을 향상시킬 수 있을 것으로 사료된다.
키워드 : 젊은 유방암, 운동, 모바일 앱, 디지털 헬스케어, 폐경 증상
학생 번호 : 2022-29148Chapter 1: Introduction.6
1.1 Significance of study
1.2 Purpose of study
1.3 Hypothesis of study
Chapter 2: Study Background.9
2.1 Breast cancer
2.2 Young breast cancer
2.3 Breast cancer and exercise
2.4 Digital health
2.5 Virtual Mate Me application
Chapter 3: Methods.29
3.1 Participants
3.2 Research design
3.3 Exercise interventions
3.4 Measurements
3.5 Statistical analysis
3.6 Ethical statements
Chapter 4: Results42
4.1 Baseline characteristics
4.2 Primary outcome: Feasibility and usability
4.3 Secondary outcome: Exercise-related effects
Chapter 5: Discussions57
5.1 Primary outcome: Feasibility and usability
5.2 Secondary outcome: Exercise-related effects
5.3 Limitations and future directions
Conclusion65
References66
Abstract in Korean78
iv
Figure List
Figure 1. Example of exercises in the Virtual Mate Me application 27
Figure 2. The feedback features in the application .28
Figure 3. Research design30
Figure 4. The System Usability Scale score..35
Figure 5. Flowchart of participant in this study.42
Figure 6. Changes in menopause symptoms from pre-test to post-test for both groups after 8
weeks of intervention.45
Figure 7. Changes in cardiorespiratory fitness from pre-test to post-test for both groups after 8
weeks of intervention.46
Figure 8. Changes in handgrip strength from pre-test to post-test for both groups after 8 weeks of
intervention47
Figure 9. Changes in lower limb strength and endurance from pre-test to post-test for both groups
after 8 weeks of intervention..49
Figure 10. Changes in shoulder ROM from pre-test to post-test for both groups after 8 weeks of
intervention 51
Figure 11. Changes in upper limb disabilities from pre-test to post-test for both groups after 8
weeks of intervention.52
Figure 12. Changes in physical activity from pre-test to post-test for both groups after 8 weeks
of intervention.53
Figure 13. Changes in quality of life from pre-test to post-test for both groups after 8 weeks of
intervention.54
Figure 14. Changes in fatigue level from pre-test to post-test for both groups after 8 weeks of
intervention55
v
Table List
Table 1. Molecular traits of breast cancer type and its subtypes..12
Table 2. Molecular traits of breast cancer subtypes17
Table 3. The phase and the duration of each session.31
Table 4. Example of exercise program in the Virtual Mate Me app32
Table 5. Rating Perceived Exertion (RPE) Scale..33
Table 6. Primary outcome variables.34
Table 7. Modified Bruce Protocol stages for treadmill test37
Table 8. The velocity and repetitions for the isokinetic test37
Table 9. Baseline characteristics of the study participants.43
Table 10. Primary outcome of the study44
Table 11. Changes in body composition from pre-test to post-test for both
groups44
Table 12. Changes in menopause symptoms from pre-test to post-test for both
groups45
Table 13. Changes in cardiorespiratory fitness from pre-test to post-test for both
groups46
Table 14. Changes in handgrip strength from pre-test to post-test for both
groups.47
Table 15. Changes in lower limb strength and endurance from pre-test to post-test for both
groups.48
Table 16. Changes in shoulder range of motion from pre-test to post-test for both
groups50
Table 17. Changes in upper limb disabilities from pre-test to post-test for both
groups.52
Table 18. Changes in physical activity from pre-test to post-test for both
groups.53
Table 19. Changes in quality of life from pre-test to post-test for both
groups54
Table 20. Changes in fatigue from pre-test to post-test for both
groups. .55
Table 21. Comparison of all variables between both groups after 8 weeks
intervention56석