• Proceedings of the Korean Society of Computer Information Conference
• /
• 2022.07a
• /
• pp.99-100
• /
• 2022

본 논문에서는 유방암 초음파 이미지를 학습한 multi-modal 구조를 이용하여 유방암을 판별하는 인공지능을 제안한다. 학습된 인공지능은 유방암을 판별과 동시에, 설명 가능한 인공지능 기법과 ROI를 함께 사용하여 종양의 위치를 나타내준다. 시각적으로 판단 근거를 제시하기 때문에 인공지능의 판단 신뢰도는 더 높아진다.

• International Journal of Advanced Culture Technology
• /
• v.5 no.1
• /
• pp.9-16
• /
• 2017

In an English multiple Wh-construction with two wh-words, only a higher-located wh-word in a sentence structure is allowed to pre-pose to the Spec of CP, which has been known as a Superiority Effect(SE). Contrary to English, Korean and Japanese have been known not to respect SE, and this phenomenon has been called an Anti-Superiority Effect(ASE). Recent studies including Takita et al(2007) and Harada(2015) among others, however, have suggested exceptional cases to ASE in Japanese, and attempted to explain them through a feature checking mechanism within the Minimalist Program(MP) by Chomsky(1995, 1998). Such explanations of MP, which are based on the abstract feature checking system, have recently been considered just as a kind of technicalia and should be backed up with further empirical evidence. In this paper, I show that there are also exceptions to ASE in Korean, and the exceptional cases to ASE both in Korean and Japanese can be well explainable based on the empirical evidence of Korean/Japanese morpho-syntax.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.32 no.1
• /
• pp.25-38
• /
• 2022

With the diversification of payment methods and games, related financial accidents are causing serious problems for users and game companies. Recently, game companies have introduced an Fraud Detection System (FDS) for game payment systems to prevent financial incident. However, FDS is ineffective and cannot provide major evidence based on judgment results, as it requires constant change of detection patterns. In this paper, we analyze abnormal transactions among payment log data of real game companies to generate related features. One of the unsupervised learning models, Autoencoder, was used to build a model to detect abnormal transactions, which resulted in over 85% accuracy. Using X-FDS (Explainable FDS) with XAI-SHAP, we could understand that the variables with the highest explanation for anomaly detection were the amount of transaction, transaction medium, and the age of users. Based on X-FDS, we derive an improved detection model with an accuracy of 94% was finally derived by fine-tuning the importance of features that adversely affect the proposed model.

• Journal of the Korean Society of Costume
• /
• v.51 no.1
• /
• pp.87-96
• /
• 2001

This study, dealing with obese bodily figures from the viewpoint of garment shape, was conducted with obese male children of late school ages(nine to eleven years old) to provide data for making ready-made clothing for obese children by directly measuring and analysing the bodily figure of the upper body of them. The results of the study are summarized as follows. 1. According to the basic statistics, for most of obese children the values of the measuring items by ages appeared to be large, and the results of comparison with the National Anthropometric Survey of Korea made in 1997 showed that the values were more large than those of standard physical conditions. 2. It was shown that in the correlation between items the upper breast circumference had a high correlation with all items, and the weight did with the following items such as circumference, height, width and thickness. 3. The results of the principal component analysis showed that six main components hand an explainable power of more than 75.60% for all materials. The first factor was thickness and obesity of the upper body, the second one the height, the third one the length of front and rear sides, the fourth one the width of neck, the fifth one the length of shoulder, and the sixth one the breast width and the wrist circumference.

• 전기의세계
• /
• v.26 no.2
• /
• pp.89-94
• /
• 1977

In order to investigate the properties of radiation resistance together with dielectric, and mechanical relaxation behaviors of polyvinylcholoride exposed to several different doses under the .gamma.-ray of Co$^{60}$ source, several observations were carried out on the exposed specimens propared by mixing dibutyl-tin-dilaulate and dibutyl-tin-dimaleate as stabilizer with or without adding dioctyl-phthalate as plasticizer. Conclusions obtained from the study are as follows: The origin of the absorption band at 1,540-1,640$cm^{-1}$ / on I.R. spectrum seems to be RCOO- ion originated from ionization of the stabilizer, and this peak can be useful as a measure of radiation resistance on polyvinylchloride. Addition of increasing plasticizer to polyvinylchloride exhibits increasing radiation resistance and the reason for this result may be attributed to aromatic resonance absorption of radiation energy by diotylphthalate. On dose dependent dielectric characteristics, nonplastized specimen shows a peak at about 10 Mrad and that this peak disappears on the plastification of specimens. Those phenomena may be explainable in considering the statistical distribution of scissored chain molecular segments as well as the plastification process of plasticizer to polyvinylchloride chain molecules.

• The Research Journal of the Costume Culture
• /
• v.8 no.5
• /
• pp.668-680
• /
• 2000

The study presents the general types and individual differences of the shape of the adult women's neck and shoulder in our country. It is based on the average value, the standard deviation, the maximum value, the standard deviation, the maximum value, the minimum value and c.v. measured from the women at the age of 19 through 64. The results of this study are as follows ; 1. The factors by factor analysis are five. The first is the size factors of the neck and shoulder. The second is shape factor of neck. The third is plane view and length factor of the neck. The fourth is the side view factor of the neck and shoulder. The fifth is the shape factor of shoulder. Therefore the shape of the neck and shoulder should be examined in the shape factors as well as in the size factors as well as in the size factors for the designing body-suitable clothes. 2. The factors of the concrete objects are the solid view of neck and shoulder, the cross-sectional view of neck line, side view of neck and shoulder and length of neck. The explainable measuring items for the factors are the difference of the length between lower neck round line and the 4㎝ above neck round line, the length of neck, the angle of inclination between fore neck and back neck, the angle of inclination of the side shoulder, the difference of the height between fore neck point and back neck point, the width of shoulder, the width of chest and the width of back. 3. The concrete objects of the neck and shoulder have five types ; Type Ⅰ is average type, Type Ⅱ is thick short neck and the drooping shoulder type, Type Ⅲ I is slender long neck and lean-back shoulder type, Type Ⅳ is thick and rising shoulder type and Type Ⅴ is slender and drooping shoulder type.

• The Bulletin of The Korean Astronomical Society
• /
• v.46 no.2
• /
• pp.74.2-75
• /
• 2021

We demonstrate that a deep learning classifier that only uses to gravitational wave (GW) detectors auxiliary channel data can distinguish various types of non-Gaussian noise transients (glitches) with significant accuracy, i.e., ≳ 80%. The classifier is implemented using the multi-scale neural networks (MSNN) with PyTorch. The glitches appearing in the GW strain data have been one of the main obstacles that degrade the sensitivity of the gravitational detectors, consequently hindering the detection and parameterization of the GW signals. Numerous efforts have been devoted to tracking down their origins and to mitigating them. However, there remain many glitches of which origins are not unveiled. We apply the MSNN classifier to the auxiliary channel data corresponding to publicly available GravitySpy glitch samples of LIGO O1 run without using GW strain data. Investigation of the auxiliary channel data of the segments that coincide to the glitches in the GW strain channel is particularly useful for finding the noise sources, because they record physical and environmental conditions and the status of each part of the detector. By only using the auxiliary channel data, this classifier can provide us with the independent view on the data quality and potentially gives us hints to the origins of the glitches, when using the explainable AI technique such as Layer-wise Relevance Propagation or GradCAM.

• Korean Journal of Veterinary Research
• /
• v.37 no.2
• /
• pp.291-299
• /
• 1997

In order to establish optimal dosage schedules and withdrawal times for sulfamethazine(SMZ) in pigs, pharmacokinetic and tissue distribution experiments were conducted in pigs. For comparative purposes, tissue depletion kinetics are also studied in rats. From three pigs administered with SMZ i.v., the pharmacokinetic profile of SMZ in two pigs was adequately described by a one-compartment open model whereas that in one pig was patterned after a two-compartment open model. Volume of distribution(Vd) was 0.48~0.57 L/kg and biological half-life($t_{1/2}$) was 11.8-16.8 h. From three pigs dosed with SMZ p.o., pharmacokinetic profile was explainable with a one-compartment open model. Time to reach maximum SMZ concentration in serum (Tmax) was 2.8 h, 3.2 h and 7.5 h. Elimination half-life was 2.8-7.5 h. The descending order in concentration of SMZ was plsama > kidney > liver > lung > heart > pancreas > spleen > duodenum > ileum > brain > adipsoe tissue from three pigs sacrificed at 5h, 29h and 54h after the administration of SMZ, p.o.. The protein binding of SMZ in pigs was 55.2%($2.5{\mu}g/ml$), 71.5% ($5{\mu}g/kg$) and 71.5%($10{\mu}g/ml$). The mean systemic bioavailability (F) of SMZ p.o. was 49.1 %. Meanwhile the pharmacokinetic profile of SMZ in rats was adequately described by a one-compartment open model. Absorption of SMZ p.o. in the rat was very rapid. In conclusion, the oral optimal dosage regimen of SMZ for pigs was the initial dose of 45.7 mg/kg followed by the maintenance dose of 30.2 mg/kg for high specific pathogens to SMZ. The time to reach below the stipulated residual allowable concentration (0.1 ppm) was calculated 93 h after oral administration of 200 mg/kg recommended by manufactureres.