• Proceedings of the Korean Information Science Society Conference
• /
• 2000.04b
• /
• pp.466-468
• /
• 2000

이 논문에서는 Helmholtz machine을 사용하여 데이터의 분포 추정을 함으로써 문서 분류기를 학습하는 방법 제안한다. Helmholtz machine 은 생성 모델과 인식 모델로 구성된 그래프 모델로서, 그래프 모델에서의 분포 추정을 보다 가능하게 하기 위한 근사 방법 중의 하나이다. Helmholtz machine에서의 각 입력 노드는 문서를 구성하는 하나의 단어에 대응하는 이진 노드이다. 입력 노드의 개수가 많아지면 그만큼 학습 시간이 증가하기 때문에, 학습 시간을 줄이면서 적정 수준의 성능을 유지하기 위해 자질 선정이 필요하다. 이러한 요구 사항을 충족시키기 위해 정보획득량(information gain)기준을 이용하였으며, 뉴스 그룹 데이터에 대해 그 성능을 측정하고 Naive Bayes를 이용한 것과 비교한다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2001.04b
• /
• pp.292-294
• /
• 2001

문서 클러스터링은 정보검색 시스템에서 검색과정의 효율성을 향상시키기 위해서 많이 사용된다. 기존의 K-means 클러스터링과 같은 거리-기반 접근 방법은 거리에 대한 척도를 정해야 하는 문제가 있고, 또한 전체 자질 공간에서 지역적 특성에 민감하기 때문에 문서 내에 노이즈가 존재할 경우 만족스러운 결과를 내지 못할 수 있다. 그리고 기본적으로 문서 데이터는 희소성(sparseness)을 가기 때문에 정규 분포를 가정한 mixture 모델을 적용하기도 어려움이 있다. 본 논문에서는 Helmoholtz machine에 의한 문서 클러스터링 방법을 제안한다. 제안되는 방법에서는 하나의 문서를 어떤 내재적인 요인(factor)들의 다양한 결합에 의한 결과로 가정하는데, 이 때의 요인은 주제어 집합 또는 적어도 의미적으로 유사한 단어들의 집합이다. 그리고 기본적으로 Helmholtz machine은 이진 데이터를 다루는데, 텍스트 문서에 나타나는 단어들의 빈도를 고려하기 위해 수정된 Helmholtz machine을 제시한다. TREC-8 adhoe 데이터와 20 Newsgroup 문서 집합에 대한 클러스터링 실험 결과, 제안된 방법이 K-means 알고리즘에 비해 우수한 성능을 보였으며 주제어 추출을 통해 문서 집합의 전체 내용 파악을 용이하게 하는 특성이 있었다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2000.04b
• /
• pp.250-252
• /
• 2000

많은 최적화 문제에서 해답들의 구조는 서로 의존성을 가지고 있다. 이러한 경우 기존의 진화연산이 사용하는 빌딩 블록 개념으로는 문제를 해결하는데 많은 어려움을 겪게 된다. 이를 극복하기 위해서 헬름홀츠 머신(Helmholtz machine)을 이용해서 데이터의 분포를 예측한 후 최적화를 수행하는 방법을 제안한다. 기존의 진화 연산을 바탕으로 하지만 교차연산이나 돌연변이 연산을 사용하는 대신에, 헬름홀츠 머신을 이용해서 데이터의 분포를 파악하고, 이를 이용해서 새로운 데이터를 생성하는 과정을 통해 최적화 과정을 수행한다. 진화연산으로 해결하는데 곤란을 겪고 있는 여러 함수들을 해결하는 이를 검증하였다.

• 전기의세계
• /
• v.29 no.1
• /
• pp.53-57
• /
• 1980

The analytical treatment for a terminal cost problem of a distributed reactor with a small singular parameter is presented. The inverse of the neutron velocity is regarded as a singular parameter, and the model, adopted for simplicity, is a cylindrically symmetrical reactor. The Helmholtz mode expension is used for the application of the optimal theory for lumped parameter systems to the spatially distributed parameter system. The closed-form solution is explicitely obtained for machine calculation.

• Journal of the Semiconductor & Display Technology
• /
• v.10 no.1
• /
• pp.33-41
• /
• 2011

Due to the recent quantum leaps forward in bio-, nano-, and information-technologies (BT, NT and IT), the precisionization and miniaturization of mechanical and electrical components are in high demand. In particular, the ITrelated equipments that take a great part in our domestic industry are in the area requiring high precision technologies. As a consequence, the researches on the development vibration isolation systems that diminish external disturbance or internal vibration are highly required. Among the components comprising the vibration isolation system, air spring has become on a focal point for the researchers due to its merits. This air spring is able to support heavy loads, keep a low natural frequency despite of having a lower value of stiffness, and control the performance of vibration isolation. However, sometimes the sole use of air spring is in demand due to some economic reasons. Under this circumstance, the damping effect of sole air spring may not enough to reduce sufficient amount of vibration. In this study, the air spring mount system connecting with an external chamber is proposed to increase or control the damping effect. To investigate its damping mechanism, the thermal and dynamic behaviors of the system is examined through a theoretical modeling approach in this part of research. In this approach, thermomechanical and Helmholtz resonator type models are to be employed for the air spring/external chambers and connecting tube system, respectively. The frequency response functions (FRFs) derived from the modeling effort are evaluated with physical parametric values and the effects of connecting tube length on these FRFs are identified through computer simulations.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 1998.03a
• /
• pp.281-286
• /
• 1998

In this paper, a Helmholtz resonator type hydraulic filter is proposed to absorb flow and pressure ripple produced from a axial piston pump. The basic principle of hydraulic filter is applied to propagation of preossure waves, reflection, absorption in cross section of discontinuity and resonance in the pipeline. This filter has advantage of the compact size and high degree of freedom of installation. The design scheme of hydraulic filter based on viscous wave theory are developed and manufactured two kinds of filter to investigate damping capability. It is experimently confirmed that these filter is absorbed to be about 20dB of flow and pressure ripple with high frequencies.

• Journal of Power System Engineering
• /
• v.12 no.4
• /
• pp.46-51
• /
• 2008

Honeycomb panel has a constructive advantage because it is constructed with a honeycomb core, so it has relatively higher strength ratio to weight. Therefore honeycomb panel has been used as the light weight panels in the high-speed railway technology and high-speed ship like as cruise yachts. Also it has been used in the aircraft and aerospace industry as a structural panel because light weight structure is indispensible in that field of industry. Recently, the honeycomb panel is embossed in the viewpoints of high oil prices as the lightweight panel of the transport machine, however the sound insulation capacity of the honeycomb panel is poorer than those of uniform and another sandwich panels. In this paper a method to improving the sound absorption coefficient of a honeycomb panel Is studied by using the Helmholtz resonator. The sound absorption coefficients for some kinds of honeycomb cores are demonstrated by the normal incident absorption coefficient method.

• Journal of the Korean Society for Industrial and Applied Mathematics
• /
• v.2 no.1
• /
• pp.27-39
• /
• 1998

We describe a fast numerical method for non-separable elliptic equations in self-adjoin form on irregular adaptive domains. One of the most successful results in numerical PDE is developing rapid elliptic solvers for separable EPDEs, for example, Fourier transformation methods for Poisson problem on a square, however, it is known that there is no rapid elliptic solvers capable of solving a general nonseparable problems. It is the purpose of this paper to present an iterative solver for linear EPDEs in self-adjoint form. The scheme discussed in this paper solves a given non-separable equation using a sequence of solutions of Poisson equations, therefore, the most important key for such a method is having a good Poison solver. High performance is achieved by using a fast high-order adaptive Poisson solver which requires only about 500 floating point operations per gridpoint in order to obtain machine precision for both the computed solution and its partial derivatives. A few numerical examples have been presented.

• Journal of KIISE:Software and Applications
• /
• v.31 no.5
• /
• pp.595-604
• /
• 2004

Automatic analysis of concepts or semantic relations from text documents enables not only an efficient acquisition of relevant information, but also a comparison of documents in the concept level. We present a multiple cause model-based approach to text analysis, where latent topics are automatically extracted from document sets and similarity between documents is measured by semantic kernels constructed from the extracted topics. In our approach, a document is assumed to be generated by various combinations of underlying topics. A topic is defined by a set of words that are related to the same topic or cooccur frequently within a document. In a network representing a multiple-cause model, each topic is identified by a group of words having high connection weights from a latent node. In order to facilitate teaming and inferences in multiple-cause models, some approximation methods are required and we utilize an approximation by Helmholtz machines. In an experiment on TDT-2 data set, we extract sets of meaningful words where each set contains some theme-specific terms. Using semantic kernels constructed from latent topics extracted by multiple cause models, we also achieve significant improvements over the basic vector space model in terms of retrieval effectiveness.