94 research outputs found
Crowdsourced Labeling for Worker-Task Specialization Model
No full textWe consider crowdsourced labeling under a d-type worker-task specialization model, where each worker and task is associated with one particular type among a finite set of types and a worker provides a more reliable answer to tasks of the matched type than to tasks of unmatched types. We design an inference algorithm that recovers binary task labels (up to any given recovery accuracy) by using worker clustering, worker skill estimation and weighted majority voting. The designed inference algorithm does not require any information about worker/task types, and achieves any targeted recovery accuracy with the best known performance (minimum number of queries per task). 11This work was supported in part by National Research Foundation of Korea under Grant 2017R1E1A1A01076340; in part by the Ministry of Science and ICT, South Korea, under the ITRC support program under Grant IITP-2021-2018-0-01402; and in part by the Institute of Information and Communications Technology Planning Evaluation (IITP) grant funded by the Korea Government MSIT under Grant 2020-0-00626
Trade-offs between Sample Complexity and Query Difficulty in Crowdsourced Data Acquisition
No full textNeurotrophin-3 regulates synapse development by modulating TrkC-PTPσ synaptic adhesion and intracellular signaling pathways
No full textNeurotrophin-3 (NT-3) is a secreted neurotrophic factor that binds neurotrophin receptor tyrosine kinase C (TrkC), which in turn binds to presynaptic protein tyrosine phosphatase σ (PTPσ) to govern excitatory synapse development. However, whether and how NT-3 cooperates with the TrkC-PTPσ synaptic adhesion pathway and TrkC-mediated intracellular signaling pathways in rat cultured neurons has remained unclear. Here, we report that NT-3 enhances TrkC binding affinity for PTPσ. Strikingly, NT-3 treatment bidirectionally regulates the synaptogenic activity of TrkC: at concentrations of 10–25 ng/ml, NT-3 further enhanced the increase in synapse density induced by TrkC overexpression, whereas at higher concentrations, NT-3 abrogated TrkC-induced increases in synapse density. Semiquantitative immunoblotting and optogenetics-based imaging showed that 25 ng/ml NT-3 or light stimulation at a power that produced a comparable level of NT-3 (6.25 μW) activated only extracellular signal-regulated kinase (ERK) and Akt, whereas 100 ng/ml NT-3 (light intensity, 25 μW) further triggered the activation of phospholipase C-γ1 and CREB independently of PTPσ. Notably, disruption of TrkC intracellular signaling pathways, extracellular ligand binding, or kinase activity by point mutations compromised TrkC-induced increases in synapse density. Furthermore, only sparse, but not global, TrkC knock-down in cultured rat neurons significantly decreased synapse density, suggesting that intercellular differences in TrkC expression level are critical for its synapse-promoting action. Together, our data demonstrate that NT-3 is a key factor in excitatory synapse development that may direct higher-order assembly of the TrkC/PTPσ complex and activate distinct intracellular signaling cascades in a concentration-dependent manner to promote competition-based synapse development processes. © 2016 the authors115141sciescopu
Feasibility study of the use of by-product iron oxide and industrial off-gas for application to chemical looping hydrogen production
No full textThe chemical looping strategy for hydrogen production (CLH2) offers a potentially viable option for efficient fuel conversion to hydrogen with the simultaneous capture of CO2. Typically, this process uses an iron-based composite as an oxygen carrier and syngas or methane as a fuel. The environmental and economic concerns motivate the use of abundant by-product iron oxide and the industrial off-gas for CLH2. Here we showed that H-2 could be simply recovered from the industrial off-gas in a circulating fluidized bed with a mixture of the inexpensive raw material of by-product iron oxide and sand particle. The fluidization of the by-product iron oxide powder, which showed poor fluidization behavior, is improved by adding 60 vol% of sand particle. The industrial off-gas was completely converted to CO2 and H2O in a two-stage fluidized mode with a solid reactant of Fe2O3 of the binary particles, and then H-2 was produced by oxidizing the reduced by-product iron oxide powder with steam. The binary particles showed consistent catalytic activity under multiple redox cycles by providing macropores with a size of 5 pm which facilitated gas diffusion. These findings provided valuable information for the future development of CLH2 based on by-products.
Locally Activating TrkB Receptor Generates Actin Waves and Specifies Axonal Fate
No full textActin waves are filamentous actin (F-actin)-rich structures that initiate in the somato-neuritic area and move toward neurite ends. The upstream cues that initiate actin waves are poorly understood. Here, using an optogenetic approach (Opto-cytTrkB), we found that local activation of the TrkB receptor around the neurite end initiates actin waves and triggers neurite elongation. During actin wave generation, locally activated TrkB signaling in the distal neurite was functionally connected with preferentially localized Rac1 and its signaling pathways in the proximal region. Moreover, TrkB activity changed the location of ankyrinG––the master organizer of the axonal initial segment-and initiated the stimulated neurite to acquire axonal characteristics. Taken together, these findings suggest that local Opto-cytTrkB activation switches the fate from minor to major axonal neurite during neuronal polarization by generating actin waves.C. 2019 Elsevier Ltd.11Nsciescopu
Noninvasive optical activation of Flp recombinase for genetic manipulation in deep mouse brain regions
No full textSpatiotemporal control of gene expression or labeling is a valuable strategy for identifying functions of genes within complex neural circuits. Here, we develop a highly light-sensitive and efficient photoactivatable Flp recombinase (PA-Flp) that is suitable for genetic manipulation in vivo. The highly light-sensitive property of PA-Flp is ideal for activation in deep mouse brain regions by illumination with a noninvasive light-emitting diode. In addition, PA-Flp can be extended to the Cre-lox system through a viral vector as Flp-dependent Cre expression platform, thereby activating both Flp and Cre. Finally, we demonstrate that PA-Flp-dependent, Cre-mediated Ca(v)3.1 silencing in the medial septum increases object-exploration behavior in mice. Thus, PA-Flp is a noninvasive, highly efficient, and easy-to-use optogenetic module that offers a side-effect-free and expandable genetic manipulation tool for neuroscience research. © The Author(s) 2019.11Nsciescopu
광 유도 신경영양물질수용체 활성의 조절 기술 개발 및 적용에 관한 연구
No full text학위논문(박사) - 한국과학기술원 : 생명과학과, 2017.2,[vii, 101 p. :]To regulate cellular functions, many biological substances like hormones, growth factors and cytokines mediate proper cellular processes. Especially in the brain, neurotrophins are well-known to be key regulators in nervous system for the cell survival, differentiation and neuronal growth. Each neurotrophin bind to its specific Trk (tropomyosin-related kinase) receptor to mediate intracellular signaling pathways.
In chapter 1, I introduce optoTrk, a new optogenetic technique to activate neurotrophin receptors with simple blue-light (488 nm) stimulation. The light-interacting module based on Arabidopsis thaliana cryptochrome 2 has homo-interacting property, used for engineering photoactivatable Trk receptors. Illumination of light efficiently activates canonical Trk downstream signaling pathways. A single light triggers transient signaling activation, reversibly reactivated with repetitive light stimuli. Moreover, the light-activated processes are tightly regulated with high spatiotemporal precision. OptoTrk system offers an effective way to control its receptor activities and experimental opportunities to study diverse biological processes both in vitro and in vivo.
In chapter 2, I introduce the importance of actin waves in neuronal polarization. By activating TrkB receptor at growth cone area specifically with optoTrkB system, I found F-actin rich waves formed at somatodendritic region and moving toward the neurite tip. Light-mediated local TrkB activation at growth cone area initiates actin wave formation and following neurite elongation. Rearrangement of actin pool was emerged from local TrkB activation to cause F-actin accumulated waves and even growth cone induction at the stimulated neurite. Additionally, this local activation of TrkB receptor at growth cone brought about accumulation of developmentally regulated protein E (drebrin E) that found mostly in growing axons. Also, axon specific proteins such as End-Binding Protein 3 (EB3) and synaptophysin changed their orientation as recruited in the TrkB activated neurite specially. Thus, local TrkB activated signaling increase F-actin rich waves from altering actin balance to determine axonal fate in neuronal development.한국과학기술원 :생명과학과
크라우드소싱을 통한 빅데이터 수집에 관한 연구: 이론적 한계 성능 및 효율적 알고리즘 성능 분석
No full text학위논문(박사) - 한국과학기술원 : 전기및전자공학부, 2023.2,[iv, 90 p. :]Today, it is crucial to collect accurate and large labeled data for artificial intelligence or machine learning algorithms to train their models. Crowdsourcing system has emerged as an effective platform to acquire labeled data with relatively low cost by using non-expert workers, because one can access to this system anytime anywhere. Although the way of crowdsourced data collection has become ubiquitous, this way can have a problem because workers who provide answers through crowdsourcing may not give accurate answers for various reasons. Therefore, it is common to infer the correct label from many answers provided by workers. However, inferring correct labels from multiple noisy answers on data has been a challenging problem, since the quality of answers varies widely across tasks and workers. Many previous works have assumed a simple model where the order of workers in terms of their reliabilities is fixed across tasks, and focused on estimating the worker reliabilities to aggregate answers with different weights. We propose a highly general -type worker-task specialization model in which the reliability of each worker can change depending on the type of a given task, where the number of types can scale in the number of tasks. In this model, we characterize the optimal sample complexity to correctly infer labels with any given accuracy, and propose an algorithm achieving the optimal result under some assumptions. We also conduct experiments both on synthetic and real datasets, and show that our algorithm outperforms the existing algorithms developed based on strict model assumptions. Finally, we conclude this dissertation by presenting a direction of future work that can be studied later.한국과학기술원 :전기및전자공학부
'은/는, 이/가, 을/를'을 중심으로
No full text학위논문(석사) -- 서울대학교대학원 : 사범대학 국어교육과(한국어교육전공), 2023. 8. 구본관.이 글은 정보구조 개념을 수용하여 한국어 조사 은/는, 이/가, 을/를의 실현변인과 기제를 표현론적 관점에서 재구성하고, 이에 대한 학습자의 인식을 학습자중심의 문법적 탐구 내용으로 제안하는 데 목적을 둔다.
그간에 한국어 조사 교육은 주로 해석론적 접근에서 수행되어 왔다. 해석론적 접근은 체계적인 문법지식을 습득하게 하고 적확한 문법 내용을 구사하게 하는 데 효율적임이 입증되어 왔으나, 모든 문법요소를 교수·학습하는 데 적합한 방식은 아니다. 특히 조사 은/는, 이/가, 을/를은 맥락에 따른 미묘한 기능 차이를 말미암아 보조사적·화용론적 기능에 대한 논의가 주의깊게 진행되어 왔던바 전통적인 조사 교수·학습으로는 충분한 습득에 한계가 있었다.
이 글에서는 은/는, 이/가, 을/를의 기능에 대한 논의와 교육이 해석론적 접근보다는 표현론적 접근에서 보다 개진될 필요성에 주목하며, 정보구조 개념을 수용하였다. 정보구조는 정보의 성격과 배열에 따른 구조적 양상으로서, 화제나 초점 등의 정보지위가 인지적·사회적 맥락에 의하여 결정된다는 점에서 표현론적 접근을 담보한다. 정보구조 개념을 수용함으로써, 조사 은/는, 이/가, 을/를이 특정한 정보와 결합하는 양상을 보이면서 학습자로 하여금 조사가 실현되는 맥락과 조사의 의미·기능을 동시에 탐구하도록 할 수 있다.
결과적으로 이 글에서는 표현론적 접근에 따라 정보에 조사가 결합되는 양상을 파악하기 위해 은/는, 이/가, 을/를이 특정한 표지임을 부정하고, 각 조사가 모두 화제·초점 등의 정보지위에 결합할 수 있는 것으로 파악하였다. 정보구조 역시 화제와 초점이 상보적인 정보지위로 이해하지 않고, 화제-평언과 초점-전제 구조가 모두 인정되는 다층위 구조를 상정하였다.
다만 조사 은/는, 이/가, 을/를은 대조의 속성원리를 갖는 것으로 이해하였다. 즉 각 조사가 텍스트에 실현될 때에는 연접 혹은 이접의 대안집합을 형성하는데. 정보지위와 유기적으로 상호작용하면서 각 대조성이 약해지거나 강해질 수 있는 것으로 파악하였다. 예컨대 조사 은/는이 비문두에 실현되거나 초점을 부여받을 때는 상승작용을 하여 배타성을 띠지만, 화제에 실현될 때는 대조원리와 길항작용을 하면서 약배타성을 띠게 된다.
한국어 학습자가 이러한 조사와 정보지위 간의 관계를 습득할 수 있도록, 정보의 성격과 배열을 고려하여 정보구조에 따른 조사 은/는, 이/가, 을/를의 실현변인과 기제 범주를 구성하였다. 이 범주의 유형은 조사 실현에 영향을 및는 실현변인의 성격에 따라 각각 언어형식적 층위와 담화인지적 층위로 구분되었다. 다시 언어형식적 층위는 무표적 문장구조·목적어 전치구조·복합문 구조로, 담화인지적 층위는 화제·초점·사건 전달방식으로 분류된다.
정보구조에 따른 조사 실현변인 및 기제 범주를 바탕으로 학습자의 조사 실현 양상과 인식을 수집하였다. 정보구조에 대한 학습자의 조사 실현양상은 국어권 학습자와 일본어권 학습자, 그리고 모어 화자를 대상으로 설문조사를 통해 수집되었고, 적절성을 기준으로 분석되었다. 한편 정보구조에 따른 조사 범주 유형에 대한 학습자의 인식은 면담을 통해 수집되었으며, 조사와 정보구조 실현기제에 대한 인식으로서 학습자가 조사를 선택하는 가운데 어떤 과정이 있었는지를 밝히기 위하여 유형화되었다. 결과적으로 한국어 학습자는 경직된 형식적 범주에 따른 학습경험, 맥락 중심의 학습 부재, 모어 간섭 등의 문제로 조사의 실현기제와 의미·기능을 제한적으로 인식하였으며, 모어 화자와 비교했을 때 조사 실현양상에서도 유의미한 차이가 나타났다.
이에 정보구조와 조사의 내적원리와 외적체계에 대한 경험을 풍부하게 하기 위한 목적으로 학습자중심의 탐구를 제안하였다. 기존의 학습자 인식의 유형은 지식의 구조 개념을 말미암아 학습자의 본유적 경험구조를 탐색할 수 있는 구체적 실체로서 이해된다. 탐구의 대상은 조사의 형태와 의미, 화용을 아우르도록 조사의 의미·기능과 연계된 담화영역과 화·청자의 인지상태를 포함하였다. 결과적으로 관찰-분석-판단이라는 단계를 거치며 학습자가 능동적으로 조사의 실현변인과 기제를 인식하게 되는 과정을 탐구의 실재로서 보였다.This paper aims to restructure the realization variables and functions of Korean postpositions 'eun/neun, i/ga, eul/reul' from a onomasiological approach, incorporating the concept of information structure. The primary objective is to propose the learners' perception of these postpositions as part of a learner-based grammatical inquiry.
Traditionally, the education of Korean postpositions has been mainly conducted from an semasiological approach. While this approach has been effective in acquiring systematic grammar knowledge and using accurate grammar structures, it might not be the most suitable method for teaching all grammar elements. Particularly, due to the subtle functional differences depending on context, postpositions 'eun/neun, i/ga, eul/reul' have limitations in being adequately acquired through traditional approaches.
In this paper, the discussion and education of the functions of 'eun/neun, i/ga, eul/reul' are highlighted from a onomasiological approach rather than an semasiological approach, recognizing the need for improvement. The concept of information structure, which refers to the structural aspects of information based on its nature and arrangement, including topic and focus, guarantees a onomasiological approach. By adopting the concept of information structure, learners can explore the context in which postpositions 'eun/neun, i/ga, eul/reul' are chosen and their meaning and functions simultaneously.
As a result, this paper grasps the combinatory aspects of postpositions with information through a onomasiological approach, denying their specific markers and acknowledging that each postposition can be combined with topic, focus, and other information positions. It also introduces a multi-layer structure that recognizes topic-predicate and focus-presupposition structures instead of viewing information structure as complementary.
However, it is understood that postpositions 'eun/neun, i/ga, eul/reul' possess a contrastive attribute principle, meaning that when realized in the text, they form alternative sets of conjunctive or disjunctive. They interact organically with information positions, leading to varying degrees of contrastiveness. For instance, when the postposition 'eun/neun' is realized in a non-main clause or receives focus, it exhibits exclusiveness, but when it is realized on the topic, it demonstrates contrastiveness through the principle of alternatives.
To help Korean language learners acquire the relationship between these postpositions and information positions, this paper constructs realization variables and mechanisms based on information structure, considering the nature and arrangement of information. The types of these variables are divided into linguistic levels, including 'non-marked sentence structure, object prepositional structure' and 'compound sentence structure' and discourse levels, including 'topic, focus' and 'event transmission methods.'
Using the foundation of information structure and the realization variables and mechanisms of postpositions, the paper collects learners' selecting patterns and perceptions of 'eun/neun, i/ga, eul/reul'. The learners' selecting patterns concerning information structure are gathered through surveys targeting Chinese learners, Japanese learners, and native speakers. The data is analyzed based on appropriateness. Additionally, learners' perceptions of postposition categories based on information structure are collected through interviews, aiming to categorize the cognitive processes involved in their choices of postpositions. Consequently, Korean language learners, due to rigid formal categories in their learning experiences, context-centered learning deficiencies, and interference from their native language, have a limited perception of postposition realization principles and meaning-functions. Comparatively, native speakers and Korean learners display significant differences in postposition realization patterns.
To address these issues and enrich learners' experiences with the inner workings and external systems of information structure and postpositions, the paper proposes a learner-based inquiry. The existing learner perception types are comprehended as concrete entities that allow exploration of learners' intrinsic experiential structures based on the concept of knowledge structure. The subject of inquiry includes the form, meaning, and pragmatics of postpositions, encompassing discourse domains and the cognitive state of speakers and listeners. Consequently, the paper reveals the process of learners actively perceiving the realization variables and functions of postpositions through the stages of observation, analysis, and judgment as the existence of inquiry.제Ⅰ 장 서론 1
제1 절 연구 목적 및 필요성 1
제2 절 선행연구 3
1. 국어학에서의 선행연구 4
2. 한국어교육에서의 선행연구 7
제3 절 연구 방법 11
제Ⅱ 장 조사 교육 연구를 위한 이론적 전제 14
제1 절 정보구조의 개념과 범위 14
1. 정보구조의 개념 14
2. 정보구조의 분절 17
3. 사건 전달방식에 따른 분절 22
제2 절 정보지위의 속성과 종류 25
1. 화제 25
1) 화제의 개념과 속성 25
2) 화제의 종류 32
2. 초점 37
1) 초점의 개념과 속성 37
2) 초점의 종류 40
제3 절 조사 교육 연구를 위한 한국어 정보구조 43
1. 정보구조에서 한국어 조사의 실현 44
1) 대조와 조사의 관계 44
2) 문장의 배열 및 구성과 정보지위의 관계 46
3) 대조와 정보지위의 관계 50
2. 정보구조에 따른 조사의 실현기제 획정 52
1) 언어형식적 층위 54
2) 담화인지적 층위 55
제Ⅲ 장 한국어 학습자의 조사 실현양상 및 인식 57
제1 절 자료 수집의 대상 및 방법 57
1. 자료 수집의 대상 57
2. 자료 분석의 방법 59
제2 절 한국어 학습자의 조사 실현양상 및 인식 63
1. 언어형식적 층위 63
1) 무표적 문장구조 조사 실현변인과 기능 인식 63
2) 목적어 전치구조 조사 실현양상과 인식 66
3) 복합문 구조 조사 살현양상과 인식 69
2. 담화인지적 층위 72
1) 화제 72
(1) 인가화제 조사 실현양상과 인식 72
(2) 대조화제 조사 실현양상과 인식 76
(3) 한정성/특정성 조사 실현양상과 인식 79
2) 초점 85
(1) 논항초점 조사 실현양상과 인식 85
(2) 초점 외 성분 조사 실현양상과 인식 87
(3) 대조초점 조사 실현양상과 인식 90
3) 사건 전달 방식 실현양상과 인식 94
제3 절 정보구조에 대한 학습자 인식의 유형과
조사 실현양상 및 인식 분석의 시사점 96
1. 정보구조에 대한 학습자 인식의 유형 96
1) 언어형식적 층위 96
(1) 통보력에 따른 대조성 인식 96
(2) 유표적 문장 구성 인식 98
2) 담화인지적 층위 99
(1) 주어짐성 인식 99
(2) 정보지위와 대조성 간의 관계 인식 102
(3) 사용역 인식 103
2. 학습 실태 조사 104
1) 한국어 교재 분석 104
2) 한국어 교수학습 실태 조사 108
3. 학습자 조사 실현양상 및 인식의 유형 분석의 시사점 110
1) 경직된 형식 범주에 따른 문제 110
2) 맥락 중심의 학습 부재 113
3) 모어 간섭 115
제Ⅳ 장 정보구조에 의한 조사 탐구의 설계 118
제1 절 학습자 중심 탐구 교육의 목적 및 구도 118
1. 학습자 중심 탐구 교육의 목적 118
2. 학습자 중심 탐구의 구도 120
제2 절 정보구조에 의한 조사 교육의 방향 및 내용 123
1. 정보구조에 의한 조사 교육의 방향 123
2. 정보구조에 의한 조사 교육의 내용 126
1) 정보구조 및 조사의 실현변인 관찰 128
2) 정보구조 및 조사의 실현기제 분석 134
3) 정보구조에 의한 조사 의미기능 판단 135
제3 절 정보구조에 의한 조사 교육의 방법 및 실재 136
1. 정보구조에 의한 조사 교육의 방법 136
2. 정보구조에 의한 조사 교육의 실재 138
제Ⅴ 장 결론 143
참고문헌 146
부록 176
Abstract 196석
- …
