http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
설명가능한 인공지능을 위한 특성기여도 분석 방법론 조사
고기혁(Gihyuk Ko),임규민(Gyumin Lim),조호묵(Homook Cho) Korean Institute of Information Scientists and Eng 2020 정보과학회논문지 Vol.47 No.12
As artificial intelligence (AI)-based technologies are increasingly being used in areas that can have big socioeconomic effects, there is a growing effort to explain decisions made by AI models. One important direction in such eXplainable AI (XAI) is the ‘feature attribution’ method, which explains AI models by assigning a contribution score to each input feature. In this work, we surveyed nine recently developed feature attribution methods and categorized them using four different criteria. Based on the categorizations, we found that the current methods focused only on specific settings such as generating local, white-box explanations of neural networks and lacked theoretical foundations such as axiomatic definitions. We suggest future research directions toward a unified feature attribution method based on our findings.
천소산(Sosan Cheon),Kenneth David Kihm,김홍구(Hong goo Kim),임규민(Gyumin Lim),박재성(Jae Sung Park),이준식(Joon Sik Lee) 대한기계학회 2015 대한기계학회 춘추학술대회 Vol.2015 No.11
Starting from graphene, “atomically thin 2-D materials” has become one of the most popular research areas for the next industrial revolution. From electronic charge carrier mobility [1] to thermal conductivity [2], the physical properties of these 2-D materials are completely different from those of bulk materials, even though made from the same atomic element. One of the most promising and feasible applications of 2-D materials is that for transparent electrodes to replace the currently adopted but relatively expensive indium tin oxide (ITO), and still the most competent candidate is graphene and its derivatives. For these applications, not only electrical but also optical characterization is critical. However, the methods to measure optical properties of graphene [3-12] are imperfect yet, in terms of the availability, robustness, and accuracy. Moreover, there is no report on the proper optical properties measurement method for doped graphene by foreign atoms. In this work, two optical measurements, surface plasmon resonance (SPR) and attenuated total reflection (ATR), have been conducted in parallel, to measure the complex refractive index (RI) of graphene layers. The graphene layers are synthesized by chemical vapor deposition (CVD) and transferred onto glass or gold substrates. Comparison of theoretical calculation results with the experimental data can determine the optical properties, i.e. complex refractive indices, of the graphene layers uniquely.
차지윤 ( Jiyoon Cha ),김인재 ( Injae Kim ),신유림 ( Yurim Shin ),임규민 ( Gyumin Lim ),윤성현 ( Sunghyun Yun ) 한국정보처리학회 2016 한국정보처리학회 학술대회논문집 Vol.23 No.2
최근 스마트기기가 발전함에 따라 종이책이 아닌 다양한 전자책뷰어가 등장하고 있다. 애플의 아이패드(iPad), 아마존의 킨들(kindle)과 같은 태블릿, 전자책 또는 스마트폰이 대표적인 예이며 스마트기기의 사용자 증가로 전용 단말기가 아닌 스마트폰 중심의 전자책 시장도 크게 성장하고 있다. 하지만 시니어 또는 스마트기기 사용을 어려워하는 사용자들이 전자책을 사용하는데 어려움을 느끼기 때문에, 전자책은 종이책에 비해 여전히 낮은 이용률을 나타내고 있다. 본 논문에서는 얼굴 표정에 따라서 전자책 설정을 자동으로 변경해 주는 방법을 제안한다. 제안한 방법은 표정감지를 위하여 OpenCV 라이브러리를 이용하여 얼굴을 검출하고, Haar-Like 피처 기법으로 사용자의 눈 모양을 검출한다. 눈이 감겨있는 경우와 찌푸린 경우를 감지하여 이에 맞게 글자크기와 화면을 자동으로 설정해 준다.