• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 1992.04a
• /
• pp.45-50
• /
• 1992

해양 가이드 타워를 대상으로 하여, 지진하중에 대한 심해응구조물의 비정상거동해법에 대하여 연구하였다. 지반운동의 비정상특성은 정상과정성분에 시간포락함수가 곱해진 형태로 모형화하였으며, 구조물의 비정상거동은 시간종속분산함수로 구하였다. 지반가속도에 대한 자기상관함수를 복소지수함수의 형태로 이상화함으로써, 구조물 거동의 시간종속함수가 해석적인 방법으로 쉽게 구할 수 있는 기법을 개발하였다. 지진의 발생시간 동안 예상되는 최대거동을 구하였으며. 이를 구조물 거동을 정상확률과정으로 가정하여 산정한 결과와 비교 분석하였다.

• Proceedings of the KIEE Conference
• /
• 2003.07d
• /
• pp.2576-2578
• /
• 2003

RBF 네트워크에서 기저함수의 위치는 네트워크의 성능에 매우 큰 영향을 미친다. 몇몇 응용들에서 교사학습을 이용한 기저함수의 위치 선정이 비교사학습에 비해 우수함을 보인다. 그러나 교사학습에 의한 네트워크는 시그모이드 네트워크와 같은 긴 학습시간을 필요로 한다. 본 논문에서는 오차함수의 gradient와 Hessian을 이용해 교사학습에서 요구하는 학습시간을 단축시키면서 기저함수의 최적위치를 추정하였다.

• Journal of The Korean Society of Grassland and Forage Science
• /
• v.11 no.4
• /
• pp.209-214
• /
• 1991

The germination of seeds is developmentally complex process requiring water uptake, which is regulated by both genotypic and environmental factors. The present study was undertaken to determine the difference in germination characteristics, and to compare the ability of the logistic and Weibull functions to describe the cumulative germination curve when two levels of osmotic potential(0, -5 bar) were put to seeds of alfalfa, tall fescue, orchardgrass, and Kentucky bluegrass. The effects of grass type, osmotic potential, and their interaction on the total germination and coefficient of germination velocity were significant(P<0.01). The Weibull equation for predicting percent cumulative germination curve of alfalfa had significantly lower residuals than the logistic equation regardless of osmotic potential(P<0.01), indicating that the Weibull equation was more efficient than the logistic equation to fit the data of the percent cumulative germination of alfalfa. The rate parameter from the logistic equation was decreased under water stress, whereas the scale and shape parameters were increased. There were significant differences in days to 20% germination estimated from the logistic and Weibull equations.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.26 no.1B
• /
• pp.27-38
• /
• 2006

The storage function model (SFM) has been used for the flood forecasting in Korea. The SFM has a simple calculation process and it is known that the model is more reasonable than linear model because it considers non-linearity of flood runoff. However, the determination of parameters is very difficult. In general, the trial and error method which is an manual calibration by the decision of a model manager. This study calibrated the parameters by the trial and error method and optimization technique. The calibrated parameters were compared with the representative parameters which are used in the Flood Control Centers in Korea. Also, the evaluation indexes on objective functions and calibration methods for the comparative analysis of simulation efficiency. As a result, the Genetic Algorithm showed the smallest variation in objective functions and, in this study, it is known that the objective function of SSR (Sum of Squared of Residual) is the best one for the flood forecasting.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.21 no.6
• /
• pp.117-122
• /
• 2021

Autonomous driving allows cars to drive without people and is being studied very actively thanks to the recent development of artificial intelligence technology. Among artificial intelligence technologies, deep reinforcement learning is used most effectively. Deep reinforcement learning requires us to build a neural network using an appropriate activation function. So far, many activation functions have been suggested, but different performances have been shown depending on the field of application. This paper compares and evaluates the performance of which activation function is effective when using deep reinforcement learning to learn autonomous driving on highways. To this end, the performance metrics to be used in the evaluation were defined and the values of the metrics according to each activation function were compared in graphs. As a result, when Mish was used, the reward was higher on average than other activation functions, and the difference from the activation function with the lowest reward was 9.8%.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.19 no.2
• /
• pp.135-141
• /
• 2019

Recently, there has been active researches about building an artificial intelligence agent that can learn how to play a game by using re-enforcement learning. The performance of the learning can be diverse according to what kinds of deep learning activation functions they used when they train the agent. This paper compares the activation functions when we train our agent for learning how to play a 2D shooting game by using re-enforcement learning. We defined performance metrics to analyze the results and plotted them along a training time. As a result, we found ELU (Exponential Linear Unit) with a parameter 1.0 achieved best rewards than other activation functions. There was 23.6% gap between the best activation function and the worst activation function.

• Journal of the Korean Data and Information Science Society
• /
• v.5 no.2
• /
• pp.45-57
• /
• 1994

이 연구에서는 Hjort(1991)에의해 제안된 누적위험률함수의 비모수적 추정량을 이용하여 무한인 구간까지 정의된 평균잔여수명함수의 추정량을 제안하고 제안된 추정량의 일치성과 점근적 정규성을 밝히고, 모의실험을 통하여 다른 추정량들과 비교하고자 한다.

• Journal of the Korea Society of Computer and Information
• /
• v.11 no.3
• /
• pp.107-115
• /
• 2006

This paper presents a new class of activation functions for Cascade Correlation learning algorithm, which herein will be called CosGauss function. This function is a cosine modulated gaussian function. In contrast to the sigmoidal, hyperbolic tangent and gaussian functions, more ridges can be obtained by the CosGauss function. Because of the ridges, it is quickly convergent and improves a pattern recognition speed. Consequently it will be able to improve a learning capability. This function was tested with a Cascade Correlation Network on the two spirals problem and results are compared with those obtained with other activation functions.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2005.11a
• /
• pp.389-392
• /
• 2005

본 논문의 컴포넌트 검색 시스템의 성능을 향상시키기 위해 사용자 피드백을 효율적으로 수행하는 방법을 제안하고자 한다. 기존의 퍼지 기법이 적용된 삼각형 모양의 퍼지화 함수는 컴포넌트를 선택할 때마다 매번 4가지 경우의 그래프를 재구성해야 하는 어려움이 있다. 본 연구에서는 이러한 피드백의 단점을 극복하기 위하여 검색된 컴포넌트의 선택여부에 따라 동일한 함수이지만 학습률을 달리할 수 있는 가우시안 함수를 이용한 상호작용 함수를 제안한다. 가우시안 함수를 피드백 함수로 채택 시 함수의 파라메타에 따른 검색 성능을 비교하고, 이를 토대로 가장 효율적인 동적 상호작용 함수를 제안하여 소수의 컴포넌트로도 좋은 검색 결과가 가능한 검색 시스템을 구축하고자 한다.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.25 no.11A
• /
• pp.1652-1660
• /
• 2000

본 논문에서는 DS/CDMA (Direct Sequence/Code Division Multipe Access) 시스템에서 임시 판정 함수로서 연판정 함수와 경판정 함수를 적용한 파이프라인화된 직렬 간섭 제어 구조(PSIC, Pipelined Successive Interference Cancellation)의 성능을 수식적으로 분석하고, 모의 실험을 통하여 검증한다. PSIC 구조는 다단 직렬 간섭 제거 구조(MSIC, Multistage Successive Interference Cancellation)가 가지는 복호지연(decoding delay)의 문제를 해결하기 위해 파이프라인 구조를 MSIC에 적용한 것이다. 제안된PSIC 구조는 하드웨어의 복잡도(hardwar complexity)를 희생하여 비트 오율(BER, Bit Error Rate)의 증가 없이 MSIC에서 발생하는 복호 지연을 줄일 수 있다. 또한 제안된 PSIC 구조에서 연판정 함수와 경판정 함수를 각 간섭 제거 단(Cancellation stage)에서의 임시 판정 함수로 사용하여 얻게 되는 PSIC 구조들의 성능을 비교한다. 분석 및 실험 결과에 의하면 제안되 PSIC 구조에서는 경판정 함수를 사용할때의 성능이 연판정 함수를 사용할때의 성능보다 우수함을 알 수 있었다.