• Journal of Internet Computing and Services
• /
• v.24 no.6
• /
• pp.1-11
• /
• 2023

Federated learning (FL) is a ground breaking machine learning paradigm that allow smultiple participants to collaboratively train models in a cloud environment, all while maintaining the privacy of their raw data. This approach is in valuable in applications involving sensitive or geographically distributed data. However, one of the challenges in FL is dealing with heterogeneous and non-independent and identically distributed (non-IID) data across participants, which can result in suboptimal model performance compared to traditionalmachine learning methods. To tackle this, we introduce FedGCD, a novel FL algorithm that employs Graph Neural Network (GNN)-based community detection to enhance model convergence in federated settings. In our experiments, FedGCD consistently outperformed existing FL algorithms in various scenarios: for instance, in a non-IID environment, it achieved an accuracy of 0.9113, a precision of 0.8798,and an F1-Score of 0.8972. In a semi-IID setting, it demonstrated the highest accuracy at 0.9315 and an impressive F1-Score of 0.9312. We also introduce a new metric, nonIIDness, to quantitatively measure the degree of data heterogeneity. Our results indicate that FedGCD not only addresses the challenges of data heterogeneity and non-IIDness but also sets new benchmarks for FL algorithms. The community detection approach adopted in FedGCD has broader implications, suggesting that it could be adapted for other distributed machine learning scenarios, thereby improving model performance and convergence across a range of applications.

• English Language & Literature Teaching
• /
• v.13 no.2
• /
• pp.155-176
• /
• 2007

An attempt to investigate the effect of gender differences on Korean students' EFL learning orientation was made. To explain a Korean EFL learning context, three criteria (cultural distance between the target country and the host country, communicative needs of the TL, the status of the TL in the host country) are adopted. Moreover, as a contrastive FL learning context from the Imposed FL learning context, a FL learning context where there is a substantial cultural distance from the TL community, communication needs of the TL do not exist, and the TL enjoys a special educational and socioeconomic status in the host country, a concept of an Integrative FL learning context is newly brought up in this paper. As the result of a questionnaire conducted in four different high schools, female learners can be claimed to be more internalized about academic and socioeconomic benefits the TL entails for their social advancement and overcoming inequality between men and women in society, albeit insignificant numerical data.

• Korean Journal of Fisheries and Aquatic Sciences
• /
• v.40 no.4
• /
• pp.248-253
• /
• 2007

We studied the maturation and spawning of female Spanish mackerel (Scomberomorus niphonius) based on 445 specimens collected monthly from January to December 2004 in the coastal waters off Busan. The fork length (FL) of S. niphonius ranged from 26.1 to 105.4 cm, the gonadosomatic index (GSI) of the females was highest in May 2004, and the spawning season take place from April to July. A significant difference was detected in the sex ratio between females and males ($X^2$ test, p<0.01). The percentage of sexually mature females exceeded 50% in the 50-60 cm (FL) size group and reached 100% in the over 70 cm (FL) size group. Fecundity (F) varied between 201,156 and 836,426 egg per female. The relationship between F and FL of the fish was expressed as $F=5.8756FL^{3.8465}$. The relationship F and body weight (BW) was expressed as F=581.421 n (BW)-4108.5. The first spawning length was 41.8 cm (FL.)

• Annual Conference on Human and Language Technology
• /
• 2021.10a
• /
• pp.311-315
• /
• 2021

근래의 자연어처리 분야에서는 잘 만들어진 도구(Library)를 이용하여 생산성 높은 개발과 연구가 활발하게 이뤄지고 있다. 이 중에 대다수는 깊은 학습(Deep-Learning, 딥러닝) 기반인데, 이런 모델들은 학습 속도가 느리고, 비용이 비싸고, 사용(Run-Time) 속도도 느리다. 이뿐만 아니라 라벨(Label)의 가짓수가 굉장히 많거나, 라벨의 구성이 단어마다 달라질 수 있는 의미분별(동형이의어, 다의어 번호 태깅) 분야에서 딥러닝은 굉장히 비효율적인 문제가 있다. 이런 문제들은 오히려 기존의 얕은 학습(Shallow-Learning)기반 모델에서는 없던 것들이지만, 최근의 연구경향에서 딥러닝 비중이 급격히 증가하면서, 멀티스레딩 같은 고급 기능들을 지원하는 얕은 학습 기반 언어모델이 새로이 개발되지 않고 있었다. 본 논문에서는 학습과 태깅 모두에서 멀티스레딩을 지원하고, 딥러닝에서 연구된 드롭아웃 기법이 구현된 자연어처리 도구인 혼합 자질 가변 표지기 ManiFL(Manifold Feature Labelling : ManiFL)을 소개한다. 본 논문은 실험을 통해서 ManiFL로 다의어태깅이 가능함을 보여주고, 딥러닝과 CRFsuite에서 높은 성능을 보여주는 개체명 인식에서도 비교할만한 성능이 나옴을 보였다.

• Biomolecules & Therapeutics
• /
• v.2 no.1
• /
• pp.82-93
• /
• 1994

DA-125, a new anthracycline antitumor antibiotic, was at dose levels of 0, 0.1, 0.3 and 1.0 mg/kg/day administered intravenously to pregnant Sprague-Dawley rats during the organogenetic period. Two-third of dams per group were subjected to caesarean section on day 20 of pregnancy and the remaining 10 dams per group were allowed to deliver. Effects of test substance on dams, embryonal development of Fl fetuses, as well as growth, behaviour and mating performance of Fl offspring were examined. 1. At 1 mg/kg, one out of the 10 dams showed difficult delivery. A decrease in food consumption, a loss in body weight and a decrease of spleen weight were found in this dose level group. At 0.3 mg/kg, difficult deliverys were observed in two out of the 10 dams. 2. At 1 mg/kg, an increased resorption rate and a decreased fetal weight were found. In addition, various types of external, visceral and skeletal malformations occurred at an incidence of 11.9, 41.8 and 14.5%, respectively. 3. At 1 mg/kg, body weight reduction, small eyeball, hydrocephalus and atrophy of sexual organs were observed in Fl offspring. One male pup receiving 0.3 mg/kg died on day 2 of lactation. The results show that the no-effect dose levels (NOELs) for dams and Fl offspring are 0.1 mg/kg/day and NOEL for Fl fetuses is 0.3 mg/kg/day.

• Journal of Microbiology and Biotechnology
• /
• v.28 no.7
• /
• pp.1141-1146
• /
• 2018

2'-Fucosyllactose (2'-FL) is one of the most important human milk oligosaccharides and has several health benefits for infants. The levels of 2'-FL in breast milk or samples from other sources can be quantified by high-performance liquid chromatography. However, this method cannot be used for simultaneous detection of the target compound in numerous samples. Here, we developed a simple method for quantifying 2'-FL in a microplate format. The method involves two steps: (i) release of $\text\tiny{L}$-fucose from 2'-FL by ${\alpha}$-(1-2,3,4,6)-$\text\tiny{L}$-fucosidase and (ii) measurement of NADPH formed during the oxidation of $\text\tiny{L}$-fucose by $\text\tiny{L}$-fucose dehydrogenase. This method enables measurement of up to 5 g/l 2'-FL in 50 min using a 96-well microplate. The efficiency and simplicity of the proposed method make it suitable for the analyses of a large number of samples simultaneously.

• Bulletin of the Korean Chemical Society
• /
• v.32 no.4
• /
• pp.1315-1320
• /
• 2011

Whey protein (WP) is a mixture of proteins, and is of high nutritional values. WP has become an important source of functional ingredients in various health-promoting foods. In this study, size-exclusion chromatography (SEC) and asymmetrical flow field-flow fractionation (AsFlFFF) were used for separation and analysis of whey proteins. It was found that a lab-prepared WP from raw milk is mostly of ${\beta}$-lactoglobulin with small amount of higher molecular weight components, while a commercial whey protein isolate (WPI) powder contains relatively larger amount of components other than ${\beta}$-lactoglobulin, including IgG and protein aggregates. Results suggest that AsFlFFF provides higher resolution for the major whey proteins than SEC in their normal operation conditions. AsFlFFF could differentiate the BSA and Albumin, despite a small difference in their molecular weights, and also was able to separate much smaller amount of aggregates from monomers. It is noted that SEC was able to show the presence of low molecular weight components other than the major whey proteins in the WP samples, which AsFlFFF could not show, probably due to the partial loss of those low molecular weight species through the membrane.

• Asian Pacific Journal of Cancer Prevention
• /
• v.16 no.7
• /
• pp.2671-2674
• /
• 2015

Thyroid cancer is the most prevalent endocrine cancer and is evident in nearly 5% of thyroid nodules. The correlation between mean platelet volume (MPV) and many other cancer types has been investigated previously. However, the correlation between papillary thyroid carcinoma (PTC) and MPV has not yet been studied in detail. The aim of this study was to examine whether MPV would be a useful inflammatory marker to differentiate PTC patients from cases of benign goiter and healthy controls. Preoperative MPV levels in patients with PTC were found to be significantly higher when compared with benign goiter patients and healthy controls ((respectively, 8.05 femtoliter (fl), 7.57 fl, 7.36 fl, p=0.001). After surgical treatment of PTC patients, a significant decrease in MPV levels was seen (8.05 fl versus 7.60 fl, p=0.005). ROC analysis suggested 7.81 as the cut-off value for MPV (AUC=0.729, sensitivity 60%, specificity 80%). In conclusion, maybe changes in MPV levels can be used as an easily available biomarker for monitoring the risk of PTC in patients with thyroid nodules, enabling early diagnosis of PTC.

• Journal of Microbiology and Biotechnology
• /
• v.18 no.8
• /
• pp.1475-1481
• /
• 2008

B3 antibody specifically binds the $Lewis^Y$-related carbohydrate antigen of many carcinomas, and it is used as a model antibody in this study. In a previous study, the Fab fragment of the antibody was fused to a 38 kDa truncated form of Pseudomonas exotoxin A, PE38, to make Fab-PE38, where PE38 is fused to the Fd fragment of the Fab domain. This parent monomer molecule, Fab-PE38, had no cysteine in the hinge region, and it could not make a disulfide bond to form a disulfide bond bridged homodimer. In this study, we constructed three different kinds of divalent Fab-toxin fusion homodimers where the toxin is fused to the light chain of Fab, $(Fab-PE38fl)_2$. In addition to the PE38 toxin fused to the light chain, these three molecules have different hinge sequences hi, h2, and h3 making Fabh1-, Fabh2-, and Fabh3-PE38fl monomers, respectively. These hinges contain only one cysteine on different positions of the hinge sequence. The disulfide bond between the hinge region of two monomers forms homodimers $(Fabh1-PE38fl)_2$, $(Fabh2-PE38fl)_2$, and $(Fabh3-PE38fl)_2$. The refolding yields of these dimers were 5-16-fold higher than a previously constructed dimer where the PE38 was fused to the Fd fragment $(Fabh2-PE38)_2$ [8]. Our data suggest that the steric repulsion between the two PE38s in $(Fabh1-PE38)_2$ during disulfide bridge formation is relieved by fusing it at the end of the light chain. The best cytotoxicity value of these dimers showed about 2.5-fold higher on an MCF7 cell line than that of the monovalent reference molecule in ng/ml scale, which is 15-fold higher in pM scale.

• Analytical Science and Technology
• /
• v.32 no.5
• /
• pp.173-184
• /
• 2019

As the size of semiconductors becomes smaller, it is necessary to perform high precision polishing of nanoscale. Ceria, which is generally used as an abrasive, is widely used because of its uniform quality, but its stability is not high because it has a high molecular weight and causes agglomeration and rapid precipitation. Such agglomeration and precipitation causes scratches in the polishing process. Therefore, it is important to accurately analyze the size distribution of ceria particles. In this study, a study was conducted to select dispersants useful for preventing coagulation and sedimentation of ceria. First, a dispersant was synthesized and a ceria slurry was prepared. The defoamer selection experiment was performed in order to remove the air bubbles which may occur in the production of ceria slurry. Dynamic light scattering (DLS) and asymmetrical flow field-flow fractionation (AsFlFFF) were used to determine the size distribution of ceria particles in the slurry. AsFlFFF is a technique for separating nanoparticles based on sequential elution of samples as in chromatography, and is a useful technique for determining the particle size distribution of nanoparticle samples. AsFlFFF was able to confirm the presence of a little quantities of large particles in the vicinity of 300 nm, which DLS can not detect, besides the main distribution in the range of 60-80 nm. AsFlFFF showed better accuracy and precision than DLS for particle size analysis of a little quantities of large particles such as ceria slurry treated in this study.