Search

Your search keyword '"Kim, So-Yun"' showing total 8,474 results
Start Over Author "Kim, So-Yun"
8,474 results on '"Kim, So-Yun"'

111. ShadowTutor: Distributed Partial Distillation for Mobile Video DNN Inference

Author

Chung, Jae-Won, Kim, Jae-Yun, and Moon, Soo-Mook

Subjects
Computer Science - Distributed, Parallel, and Cluster Computing
Abstract

Following the recent success of deep neural networks (DNN) on video computer vision tasks, performing DNN inferences on videos that originate from mobile devices has gained practical significance. As such, previous approaches developed methods to offload DNN inference computations for images to cloud servers to manage the resource constraints of mobile devices. However, when it comes to video data, communicating information of every frame consumes excessive network bandwidth and renders the entire system susceptible to adverse network conditions such as congestion. Thus, in this work, we seek to exploit the temporal coherence between nearby frames of a video stream to mitigate network pressure. That is, we propose ShadowTutor, a distributed video DNN inference framework that reduces the number of network transmissions through intermittent knowledge distillation to a student model. Moreover, we update only a subset of the student's parameters, which we call partial distillation, to reduce the data size of each network transmission. Specifically, the server runs a large and general teacher model, and the mobile device only runs an extremely small but specialized student model. On sparsely selected key frames, the server partially trains the student model by targeting the teacher's response and sends the updated part to the mobile device. We investigate the effectiveness of ShadowTutor with HD video semantic segmentation. Evaluations show that network data transfer is reduced by 95% on average. Moreover, the throughput of the system is improved by over three times and shows robustness to changes in network bandwidth., Comment: Accepted at ICPP 2020

Published
2020
Full Text
View/download PDF

115. Variations in medical practice of retinopathy of prematurity among 8 Asian countries from an international survey

Author

Youn, Young-Ah, Kim, Sae Yun, Cho, Su Jin, Chang, Yun Sil, Miyake, Fuyu, Kusuda, Satoshi, Iskandar, Adhi Teguh Perma, Rohsiswatmo, Rinawati, Dewi, Rizalya, Chee, Seok Chiong, Neoh, Siew Hong, Imperial, Ma. Lourdes S., Velasco, Belen Amparo E., Quek, Bin Huey, Lin, Yuh-Jyh, Chang, Jui-Hsing, Nuntnarumit, Pracha, Ngerncham, Sopapan, Supapannachart, Sarayut, Ozawa, Yuri, Tomotaki, Seiichi, Prempunpong, Chatchay, Prempraphan, Pathaporn, and Isayama, Tetsuya

Published
2023
Full Text
View/download PDF

133. Development of Novel-Engineering-Based Maker Education Instructional Model

Author

Kim, Ji-Yun, Seo, Jae Seon, and Kim, Kwihoon

Abstract

Attaining equitable education for all is one of the sustainable development goals. Not only the gap of education between urban and rural students but also the gap of access to information and communications technology also exists. Thus, an effort to resolve this gap is necessary. This study combines novel engineering, which is the fusion of engineering and literacy educational methods with maker education, to create a sustainable model of maker education. Through this process, an instructional model for a novel engineering-based maker education is proposed that would enable the utilization of the model in rural elementary school environment. To achieve this, this study analyzed the literature regarding maker education, novel engineering, and design thinking to develop, draft, and verify internal validity through an expert's review and usability evaluation. In addition, this study includes an example class for rural elementary students with application of the model to verify the model's external validity and confirm that the model significantly improved the maker mindset. Finally, this study proposes final instructional model by modifying and improving the draft based on the recommendations derived from the validity verification process. The proposed instructional model was developed as a suitable form for rural school students, which can use low-cost teaching tools to lower the entry barrier to information and communications technology education. The model process developed in this study has the advantage of raising convergence problem-solving abilities and the maker mindset, in terms of a making project that fuses literature and engineering, among students.

Published
2022
Full Text
View/download PDF

140. Psychometric Properties of an English Writing Self-Efficacy Scale: Aspects of Construct Validity

Author

Sun, Ting, Wang, Chuang, and Kim, Stella Yun

Abstract

This study provided validity evidence of the Questionnaire of English Writing Self-Efficacy (QEWSE) based on the five sources of validity evidence described in the "Standards for Educational and Psychological Testing" (AERA, APA, & NCME, 2014). Items of this measure demonstrated close alignment with the linguistic domain and the writing self-efficacy and self-regulation theories, providing validity evidence based on test content. Confirmatory factor analysis revealed a satisfactory global fit of the four-factor (ideation, skills, use, and self-regulation) model to the data, providing internal structure-based evidence. Significant correlations between self-efficacy ratings, writing proficiency, and writing self-regulated learning strategies supported the validity evidence based on relations to other variables. The reliability evidence was evaluated with consistencies of participants' responses to items and further supported using Item Response Theory (IRT). IRT analyses suggested that measurement precision was greatest at the lower range of the latent trait scale, which provided justification for the primary use of the instrument to identify students with low writing self-efficacy.

Published
2022
Full Text
View/download PDF

146. Ethanos: Lightweight Bootstrapping for Ethereum

Author

Kim, Jae-Yun, Lee, Jun-Mo, Koo, Yeon-Jae, Park, Sang-Hyeon, and Moon, Soo-Mook

Subjects
Computer Science - Computation and Language, Computer Science - Distributed, Parallel, and Cluster Computing
Abstract

As ethereum blockchain has become popular, the number of users and transactions has skyrocketed, causing an explosive increase of its data size. As a result, ordinary clients using PCs or smartphones cannot easily bootstrap as a full node, but rely on other full nodes such as the miners to run or verify transactions. This may affect the security of ethereum, so light bootstrapping techniques such as fast sync has been proposed to download only parts of full data, yet the space overhead is still too high. One of the biggest space overhead that cannot easily be reduced is caused by saving the state of all accounts in the block's state trie. Fortunately, we found that more than 90% of accounts are inactive and old transactions are hard to be manipulated. Based on these observations, this paper propose a novel optimization technique called ethanos that can reduce bootstrapping cost by sweeping inactive accounts periodically and by not downloading old transactions. If an inactive account becomes active, ethanos restore its state by running a restoration transaction. Also, ethanos gives incentives for archive nodes to maintain the old transactions for possible re-verification. We implemented ethanos by instrumenting the go-ethereum (geth) client and evaluated with the real 113 million transactions from 14 million accounts between 7M-th and 8M-th blocks in ethereum. Our experimental result shows that ethanos can reduce the size of the account state by half, which, if combined with removing old transactions, may reduce the storage size for bootstrapping to around 1GB. This would be reasonable enough for ordinary clients to bootstrap on their personal devices.

Published
2019

147. Magnetic Domain Wall Motion due to AC Bias-Driven Resonances

Author

Kim, Duck-Ho, Kim, Dong-Hyun, Kim, Dae-Yun, Choe, Sug-Bong, Ono, Teruo, Lee, Kyung-Jin, and Kim, Se Kwon

Subjects
Condensed Matter - Mesoscale and Nanoscale Physics
Abstract

Most of the existing researches on the dynamics of a domain wall (DW) have focused on the effect of DC biases, where the induced velocity is determined by the bias strength. Here we show that AC biases such as a field or a current are also able to move a DW via synchronization between the DW angle and the phase of the AC bias. The resulting DW velocity is proportional to the driving frequency of the AC bias, but independent of the bias strength, offering potentially low-power operations of DW devices. The AC-bias-driven DW motion is shown to exhibit a phase locking-unlocking transition, a critical phenomenon akin to the Walker breakdown of a DC-bias-driven DW motion. Our work shows that a DW can be driven resonantly by synchronizing its angle to AC biases, shedding a light on hitherto overlooked utility of internal degree of freedom for driving magnetic textures., Comment: 20 pages, 4 figures

Published
2019

148. Endogenous TLR2 ligand embedded in the catalytic region of human cysteinyl-tRNA synthetase 1

Author

Cho, Seongmin, Kim, Sang Bum, Lee, Youngjin, Song, Ee Chan, Kim, Uijoo, Kim, Hyeong Yun, Suh, Ji Hun, Goughnour, Peter C, Kim, YounHa, Yoon, Ina, Shin, Na Young, Kim, Doyeun, Kim, Il-Kyu, Kang, Chang-Yuil, Jang, Song Yee, Kim, Myung Hee, and Kim, Sunghoon

Subjects
Biomedical and Clinical Sciences, Oncology and Carcinogenesis, Immunology, Vaccine Related, Cancer, Biotechnology, Prevention, Immunization, Development of treatments and therapeutic interventions, Underpinning research, 5.1 Pharmaceuticals, 1.1 Normal biological development and functioning, Inflammatory and immune system, Good Health and Well Being, Amino Acyl-tRNA Synthetases, Animals, Cancer Vaccines, Catalytic Domain, Dendritic Cells, Female, Humans, Immunity, Cellular, Immunotherapy, Ligands, Mice, Mice, Inbred C57BL, Neoplasms, Experimental, T-Lymphocytes, Cytotoxic, Toll-Like Receptor 2, adjuvants, immunological, immunology, oncology, vaccination, Oncology and carcinogenesis
Abstract

The generation of antigen-specific cytotoxic T lymphocyte (CTL) responses is required for successful cancer vaccine therapy. In this regard, ligands of Toll-like receptors (TLRs) have been suggested to activate adaptive immune responses by modulating the function of antigen-presenting cells (APCs). Despite their therapeutic potential, the development of TLR ligands for immunotherapy is often hampered due to rapid systemic toxicity. Regarding the safety concerns of currently available TLR ligands, finding a new TLR agonist with potent efficacy and safety is needed. A unique structural domain (UNE-C1) was identified as a novel TLR2/6 in the catalytic region of human cysteinyl-tRNA synthetase 1 (CARS1) using comprehensive approaches, including RNA sequencing, the human embryonic kidney (HEK)-TLR Blue system, pull-down, and ELISA. The potency of its immunoadjuvant properties was analyzed by assessing antigen-specific antibody and CTL responses. In addition, the efficacy of tumor growth inhibition and the presence of the tumor-infiltrating leukocytes were evaluated using E.G7-OVA and TC-1 mouse models. The combined effect of UNE-C1 with an immune checkpoint inhibitor, anti-CTLA-4 antibody, was also evaluated in vivo. The safety of UNE-C1 immunization was determined by monitoring splenomegaly and cytokine production in the blood. Here, we report that CARS1 can be secreted from cancer cells to activate immune responses via specific interactions with TLR2/6 of APCs. A unique domain (UNE-C1) inserted into the catalytic region of CARS1 was determined to activate dendritic cells, leading to the stimulation of robust humoral and cellular immune responses in vivo. UNE-C1 also showed synergistic efficacy with cancer antigens and checkpoint inhibitors against different cancer models in vivo. Further, the safety assessment of UNE-C1 showed lower systemic cytokine levels than other known TLR agonists. We identified the endogenous TLR2/6 activating domain from human cysteinyl-tRNA synthetase CARS1. This novel TLR2/6 ligand showed potent immune-stimulating activity with little toxicity. Thus, the UNE-C1 domain can be developed as an effective immunoadjuvant with checkpoint inhibitors or cancer antigens to boost antitumor immunity.

Published
2020

Catalog

Books, media, physical & digital resources
See catalog results