http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
- 中文 을 입력하시려면 zhongwen을 입력하시고 space를누르시면됩니다.
- 北京 을 입력하시려면 beijing을 입력하시고 space를 누르시면 됩니다.
RISS 인기검색어
- 검색키워드
- 저자 : Frank Chongwoo Park (검색결과 7 건)
- 무료
- 기관 내 무료
- 유료
-
Some Fundamental Problems in Robotics
Frank Chongwoo Park 한국산업응용수학회 2005 한국산업응용수학회 학술대회 논문집 Vol.- No.-
We review some fundamental problems that arise in robot mechanics, and show how tools from differential geometry and Lie group theory have played an integral part in their solution.
-
Randomized path planning on vector fields
Ko, Inyoung,Kim, Beobkyoon,Park, Frank Chongwoo SAGE Publications 2014 The International journal of robotics research Vol.33 No.13
<P>Given a vector field defined on a robot’s configuration space, in which the vector field represents the system drift, e.g. a wind velocity field, water current flow, or gradient field for some potential function, we present a randomized path planning algorithm for reaching a desired goal configuration. Taking the premise that moving against the vector field requires greater control effort, and that minimizing the control effort is both physically meaningful and desirable, we propose an integral functional for control effort, called the <B>upstream criterion</B>, that measures the extent to which a path goes against the given vector field. The integrand of the upstream criterion is then used to construct a rapidly exploring random tree (RRT) in the configuration space, in a way such that random nodes are generated with an a priori specified bias that favors directions indicated by the vector field. The resulting planning algorithm produces better quality paths while preserving many of the desirable features of RRT-based planning, e.g. the Voronoi bias property, computational efficiency, algorithmic simplicity, and straightforward extension to constrained and nonholonomic problems. Extensive numerical experiments demonstrate the advantages of our algorithm vis-à-vis existing optimality criterion-based planning algorithms.</P>
-
-
Model Dependency of the Performance in Generative Local Metric Learning
Yung-Kyun Noh,Frank Chongwoo Park,Daniel D. Lee 한국정보과학회 2011 한국정보과학회 학술발표논문집 Vol.38 No.2B
Nearest neighbor classification with generative local metric (GLM) learning is a hybrid method of the discriminative and generative approaches. A discriminative nearest neighbor classifier does not consider any model of data, while a generative classifier unavoidably adopts a particular form of the probability density function. In this work, we illuminate how these discriminative and generative approaches have different advantages and show how the advantages of both can be resolved into a GLM method. We present various examples that clearly show the different regimes where the discriminative and generative approaches should outperform each other. In these examples, we show that the GLM is robust to the usage of an incorrect model, enhancing nearest neighbor classifier performance even when the model is not exact.
-
Fluid Dynamic Models for Bhattacharyya-Based Discriminant Analysis
Noh, Yung-Kyun,Hamm, Jihun,Park, Frank Chongwoo,Zhang, Byoung-Tak,Lee, Daniel D. IEEE 2018 IEEE transactions on pattern analysis and machine Vol.40 No.1
<P>Classical discriminant analysis attempts to discover a low-dimensional subspace where class label information is maximally preserved under projection. Canonical methods for estimating the subspace optimize an information-theoretic criterion that measures the separation between the class-conditional distributions. Unfortunately, direct optimization of the information-theoretic criteria is generally non-convex and intractable in high-dimensional spaces. In this work, we propose a novel, tractable algorithm for discriminant analysis that considers the class-conditional densities as interacting fluids in the high-dimensional embedding space. We use the Bhattacharyya criterion as a potential function that generates forces between the interacting fluids, and derive a computationally tractable method for finding the low-dimensional subspace that optimally constrains the resulting fluid flow. We show that this model properly reduces to the optimal solution for homoscedastic data as well as for heteroscedastic Gaussian distributions with equal means. We also extend this model to discover optimal filters for discriminating Gaussian processes and provide experimental results and comparisons on a number of datasets.</P>
-
노영균(Yung-Kyun Noh),박종우(Frank Chongwoo Park),다니엘 리(Daniel D. Lee) 한국정보과학회 2012 정보과학회논문지 : 소프트웨어 및 응용 Vol.39 No.5
생성국부거리(Generative Local Metric, GLM) 학습을 통한 최근린 분류(Nearest Neighbor Classification, NN Classification)는 생성적(generative) 접근방법과 분류적(discriminative) 접근방법의 합성 모델이다. 두 방법의 차이는 분류형 형태를 띄는 최근린 분류기가 데이터의 모델을 고려하지 않는 반면, 생성형 분류기는 언제나 확률 밀도 함수의 형태를 가정하고 학습하게 된다는 데 있다. 이 연구에서는 어떻게 이러한 서로 다른 분류적, 생성적 접근이 서로 다른 종류의 강점을 가지게 되고, 어떻게 GLM은 이러한 양쪽의 강점을 모두 취하게 되는지 설명한다. 다양한 예시를 통해 우리는 어떠한 경우에 생성적 방법이 분류적 방법보다, 혹은 분류적 방법이 생성적 방법보다 좋은 성능을 보이게 되는가를 알아보고, 이러한 경우에서 비록 모델이 정확하지 않은 경우에 대해서도 GLM이 원래의 최근린 분류 성능을 계속 향상 시키는 좋은 특징이 있는 것을 보인다. Nearest neighbor (NN) classification with generative local metric (GLM) learning is a hybrid method of the discriminative and generative approaches. A discriminative NN classifier does not consider any model of data, while a generative classifier unavoidably adopts a particular form of the probability density function. In this work, we illuminate how these discriminative and generative approaches have different advantages and show how the advantages of both can be resolved into a GLM method. We present various examples that clearly show the different regimes where the discriminative and generative approaches should outperform each other. In these examples, we show that the GLM is robust to the usage of an incorrect model, enhancing NN classifier performance even when the model is not exact.
-
Classification of impinging jet flames using convolutional neural network with transfer learning
Minwoo Lee,Sangwoong Yoon,Ju Han Kim,Yuangang Wang,Kee-Man Lee,Frank Chongwoo Park,Chae Hoon Sohn 대한기계학회 2022 JOURNAL OF MECHANICAL SCIENCE AND TECHNOLOGY Vol.36 No.3
Depending on the equivalence ratio and the Reynolds number, impinging jet flames exhibit several modes of thermoacoustic oscillation. In this study, we present a machine-learning-based method for classifying the regimes of thermoacoustic oscillation. We perform transfer learning to train the convolutional neural network model designed to classify flame images. We show that an accurate classification of impinging jet flames is achieved with an accuracy of 93.6 % by using just a single snapshot image. This study constitutes the first demonstration of transfer learning in classifying fluid images, opening up new possibilities for robust image-based diagnostics of various fluid and combustion systems.
연관 검색어 추천
이 검색어로 많이 본 자료
활용도 높은 자료
