http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
GAN 을 이용한 유전체 발현 프로필 생성에서 활성화 함수 별 성능 비교
Suhan Son,Junhee Seok 한국차세대컴퓨팅학회 2021 한국차세대컴퓨팅학회 학술대회 Vol.2021 No.05
딥러닝은 비선형 데이터를 다룰 수 있으면서, 내부적인 특성들과 연관 관계 등을 학습할 수 있다는 점에서 많은 분야에서 쓰이고 있다. 딥러닝이 비선형 데이터를 다루는 것이 가능한 이유는 활성화 함수가 사용되기 때문이다. 활성화 함수로는 비선형 함수를 이용하여 입력과 결과가 비선형성을 띌 수 있게 된다. 그런데 활성화 함수의 종류는 다양하며 각각 특성이 다르다. 따라서 어떤 활성화 함수가 사용이 되는지에 따라 결과의 차이가 발생할 수 있다. 본 연구에서는 활성화 함수의 차이가 딥러닝 학습에서 영향을 미치는 지에 대해 알아보고자 한다. 먼저 딥러닝 모델 중 하나인 GAN을 세포의 유전자 데이터를 이용하여 여러 모델 학습을 시킨다. 그 과정에서 각 GAN 모델의 생성자와 식별자는 다른 모델의 활성화 함수와 다른 조합으로 학습된다. 그 후 서로 다른 활성화 함수 조합으로 학습된 GAN의 생성자를 통해 가짜 데이터를 생성한다. 최종적으로 각 모델이 만들어 낸 가짜 데이터와 실제 데이터와의 상관 분석을 통해서 생성된 모델 간의 성능을 비교하여, 활성화 함수의 차이가 딥러닝 학습에 영향을 미치는 지 확인한다.
Suhan Choi,Ji-Hoon Yun,Kae Won Choi IEEE 2013 IEEE COMMUNICATIONS LETTERS Vol.17 No.6
<P>In this letter, functional duality between Distributed Source Coding with One Distortion Criterion (DSC-ODC) and correlated messages and Semi-Deterministic Broadcast Channel Coding (SD-BCC) with correlated messages is considered. It is shown that under certain conditions, for a given DSC-ODC problem with correlated messages, a functional dual SD-BCC problem with correlated messages can be obtained, and vice versa. In particular, the correlation structure of the messages in the two dual problems are the same. The source distortion measure and the channel cost measure for this duality are also specified.</P>
Effect of Beverage Containing Fermented Akebia quinata Extracts on Alcoholic Hangover
Suhan Jung,Sang Hoon Lee,Young Sun Song,Seo Yeon Lee,So Young Kim,Kwang Suk Ko 한국식품영양과학회 2016 Preventive Nutrition and Food Science Vol.21 No.1
The present study was conducted to investigate the effects of beverages containing fermented Akebia quinata extracts on alcoholic hangover. For this study, 25 healthy young men were recruited. All participants consumed 100 mL of water (placebo), commercial hangover beverage A or B, fermented A. quinata leaf (AQL) or fruit (AQF) extract before alcohol consumption. After 1 h, all participants consumed a bottle of Soju, Korean distilled liquor (360 mL), containing 20% alcohol. Blood was collected at 0 h, 1 h, 3 h, and 5 h after alcohol consumption. The plasma alanine transaminase (ALT) activity was highest in the placebo group. Compared with the control group, the AQL and AQF groups showed decreased ALT activity at 5 h after alcohol consumption. Plasma ethanol concentration was increased after alcohol intake and peaked at 3 h after alcohol consumption. Compared with the control group, the A group showed a higher plasma ethanol concentration at 1 h (P<0.05). At 3 h after alcohol consumption, the AQF group showed the lowest mean plasma ethanol concentration compared to the other groups; however, there were no statistical differences. After 5 h of alcohol consumption, the AQL and AQF groups showed lower plasma ethanol concentrations compared with the B group. The sensory evaluation score for the fermented A. quinata fruit extract was lower than for the commercial hangover beverages. In conclusion, the present intervention study results suggest that fermented A. quinata extracts alleviate alcoholic hangover and reduce plasma ethanol concentrations.
A study on test method for evaluation of mobile robot trajectory estimation
Suhan Lee,Seungsub Oh 제어로봇시스템학회 2021 제어로봇시스템학회 국제학술대회 논문집 Vol.2021 No.10
Indoor mobile robot location is the most important technology for service robots. Conventionally, there are various methods for estimating the robot trajectory to evaluate performance of robot localization. Since the operating environment and purpose of the robot are diverse, an appropriate method must be selected for the testing. In this paper, we present a study comparing the trajectory estimation methods of mobile robots using internal and external sensors to assist in estimating a suitable method for evaluation.
NODEIK: Solving Inverse Kinematics with Neural Ordinary Differential Equations for Path Planning
Suhan Park,Mathew Schwartz,Jaeheung Park 제어로봇시스템학회 2022 제어로봇시스템학회 국제학술대회 논문집 Vol.2022 No.11
This paper proposes a novel inverse kinematics (IK) solver of articulated robotic systems for path planning. IK is a traditional but essential problem for robot manipulation. Recently, data-driven methods have been proposed to quickly solve IK for path planning. These machine learning-based models can handle a large amount of IK requests at once by leveraging the GPU. However, such methods suffer from reduced accuracy and considerable training time. We propose an IK solver that improves accuracy and memory efficiency with continuous normalizing flows by utilizing the continuous hidden dynamics of a Neural ODE network. The performance is compared using multiple robots, and our method is shown to be highly performant on complex (including dual end effector) manipulators.