http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
비침습적 생체공학 방법으로 측정한 정상 몽골인 피부의 생리적 특성
나르만다흐 ( Narmandakh Jugee ),전혜찬 ( Hye Chan Jeon ),연제호 ( Je Ho Yeon ),백승환 ( Seung Hwan Paik ),최재우 ( Jae Woo Choi ),박영운 ( Young Woon Park ),은희철 ( Hee Chul Eun ),권오상 ( Oh Sang Kwon ),서정선 ( Jeong Sun Seo ) 대한피부과학회 2012 大韓皮膚科學會誌 Vol.50 No.5
Background: Physiological parameters of the skin measured by non-invasive methods have been considerably developed. It is known that there are some differences in physiologic parameters between different races. Objective: The purpose of this study is to understand the differences between the races. Methods: A total of 757 Mongolian participated in this study. All subjects had no major history of skin diseases requiring medical treatment. Several instruments were used such as Sebumeter, Corneometer, Mexameter for measuring sebum excretion rate, values of capacitance, melanin index and erythema index respectively. These were measured on various areas of the skin such as cheek, forehead, palm, outer arm, inner arm, back and buttock. Results: The sebum excretion rates showed higher in males than in females on the cheek and forehead in Mongolian. However, they showed higher in the females on the palm. There were good correlation between the skin sebum level and the capacitance in males and females. The melanin index and erythema index showed higher in males than in females at all sites. When we compared this with the data published in other countries, many data were similar to Korean rather than a western country`s. However, the capacitance data were lower than those of Korean. Conclusion: Although, this study has some limit for direct comparison between each race, our results can be used as basic data for the comparison between Mongolian and other races in the future. (Korean J Dermatol 2012;50(5):413∼418)
비대칭 데이터셋 기반 딥러닝을 위한 적응적 학습률 조절 기법
박태우(Tae Woo Park),양현종(Hyun Jong Yang) 한국통신학회 2021 한국통신학회 학술대회논문집 Vol.2021 No.2
본 논문은 비대칭 데이터셋 기반 딥러닝을 다룬다. 많은 클래스 분류 문제는 특정 데이터 수집의 어려움으로 비대칭 데이터셋을 가진다. 데이터량의 심각한 비대칭으로 인한 과편향 문제를 막기 위해 기존의 과소 표본, 과대 표본을 사용하지 않고 클래스별로 학습률을 적응적으로 조절하는 기법을 제안한다. 실험에서 이 기법은 Random undersampling 보다 10% 높은 정확도를 보였고 원본데이터보다 높은 F1 score 를 보이며 undersampling 의 비대칭 문제 해결의 장점과 원본데이터의 데이터 손실이 없다는 장점을 동시에 가진다.