• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2003.09a
• /
• pp.162-165
• /
• 2003

Differential learning relies on the differentiated values of nodes, whereas the conventional learning depends on the values themselves of nodes. In this paper, I elucidate the differential learning in the framework maximum likelihood learning of linear generative model with latent variables obeying random walk. I apply the idea of differential learning to the problem independent component analysis(ICA), which leads to differential ICA. Algorithm derivation using the natural gradient and local stability analysis are provided. Usefulness of the algorithm is emphasized in the case of blind separation of temporally correlated sources and is demonstrated through a simple numerical example.

• The Journal of the Acoustical Society of Korea
• /
• v.25 no.4E
• /
• pp.148-153
• /
• 2006

A new normalized MBD algorithm is presented for nonstationary convolutive mixtures and its properties/modifications are discussed in details. The proposed algorithm normalizes the signal spectrum in the frequency domain to provide faster stable convergence and improved separation without whitening effect. Modifications such as nonholonomic constraints and off-diagonal learning to the proposed algorithm are also discussed. Simulation results using a real-world recording confirm superior performanceof the proposed algorithm and its usefulness in real world applications.

• Proceedings of the KIEE Conference
• /
• 1998.11b
• /
• pp.660-662
• /
• 1998

데이터 sphering은 신호처리 및 신경망 분야에서 널리 이용되는 기본적인 처리과정으로, 그 목적은 주어진 데이터간에 correlation을 제거하는 것이다. 본 논문에서는 회귀 신경망을 도입하여, 데이터 sphering을 위한 새로운 on-line 알고리듬을 제시한다. 정보 이론에 입각한 risk 함수와, 최근에 제시된 natural gradient을 이용하여 새로운 데이터 sphering 알고리듬을 유도한다. 새로 제시된 알고리듬의 성능을 기존에 널리 이용되어 온 anti-Hebbian 학습 알고리듬과 비교 분석한다.

• Proceedings of the Acoustical Society of Korea Conference
• /
• spring
• /
• pp.41-44
• /
• 2002

Blind source separation (BSS) is a fundamental problem that is encountered in many practical applications. In most existing methods, stationary sources are considered higher-order statistics is necessary either explicitly or implicitly. But, many natural signals are nonstationary, and it is possible to perform BSS using only second-order statistics. Our method is based on only second order statistics. The algorithms are developed using the gradient descent method in orthogonality constraint and their performance is confirmed by numerical experiments.

• Structural Engineering and Mechanics
• /
• v.77 no.5
• /
• pp.587-599
• /
• 2021

Based on Euler-Bernoulli beam theory and continuous element method, the free vibration of bi-dimensional functionally graded beams is investigated. It is assumed that the material properties vary exponentially along the beam thickness and length. The characteristic frequency equations of beams with different boundary conditions are obtained by transfer matrix method. The validity of the proposed method is assessed through comparison with available results. Parametric studies are carried out to analyze the influences of the gradient indexes and the beam slenderness ratio on the natural frequencies of bi-dimensional functionally graded beams.

• Journal of the Korean Mathematical Society
• /
• v.58 no.5
• /
• pp.1279-1298
• /
• 2021

The aim of this paper is to prove the existence of the global attractor of the Cauchy problem for a semilinear degenerate hyperbolic equation involving strongly degenerate elliptic differential operators. The attractor is characterized as the unstable manifold of the set of stationary points, due to the existence of a Lyapunov functional.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2007.04a
• /
• pp.33-36
• /
• 2007

최근에 인공지능 분야에서는, 국내외적으로 강화학습(reinforcement learning)에 관한 관심이 크게 증폭되고 있다. 강화학습의 최근 경향을 살펴보면, 크게 가치함수를 직접 활용하는 방법(value function-based methods), 제어 전략에 대한 탐색을 활용하는 방법(policy search methods), 그리고 액터-크리틱 방법(actor-critic methods)의 세가지 방향으로 발전하고 있음을 알 수 있다. 본 논문에서는 이중 세 번째 부류인 액터-크리틱 방법 중 NAC(natural actor-critic) 기법의 한 종류인 RLS-NAC(recursive least-squares based natural actor-critic) 알고리즘을 다양한 트레이스 감쇠계수를 사용하여 연속제어입력(real-valued control inputs)으로 제어되는 Kimura의 기는 로봇에 대해 적용해보고, 그 성능을 기존의 SGA(stochastic gradient ascent) 알고리즘을 이용하여 학습한 경우와 비교해보도록 한다.

• Journal of the Korean Crystal Growth and Crystal Technology
• /
• v.7 no.1
• /
• pp.41-46
• /
• 1997

The change of flow field and the effects of convective heat transfer on the shape and location of melt/crystal interface has been studied during the crystal growth by the heat exchanger method. Although the thermal structure is stable in the crucible, the flow due to the natural convection driven by radial temperature gradient is significant, because the thermal stability is broken by the hemispherical melt/crystal interface shape. The maximum interface deflection with convection is smaller than without and the convective heat transfer should be considered to simulate the heat transfer process of heat exchanger method rigorously.

• Journal of the Korean Institute of Landscape Architecture
• /
• v.18 no.2
• /
• pp.95-110
• /
• 1990

Wang -gok Villages was forming the space marking the centripetal system from frame composed of adjacent mountains and the cosy and comfortable space perccplively and psychologically, and both the openness and closeness existed together The analysis of the circulation structure in the exterior spaces of home showed that generally one must pass through kitchen to approach to the rear yard having the strong decree of enclosure and privacy. The determinants of the space organization in Wang-gok villages and the exterior of its homes were direction and gradient as natural factor, and traditional thought, action and social class as sococultural factors. The configuration of rear garden was classified into flat type, terrace type and mixed type of natural hill and terrace. The rear garden was created purposely and actively, emphasizing the harmony with and approach to nature.

• Particle and aerosol research
• /
• v.10 no.3
• /
• pp.99-108
• /
• 2014

Liquid filtration by membrane filters is essential for the preparation of ultrapure water in semiconductor manufacturing processes. The separation of submicrometer particles suspended in ultrapure water with a laminated fibrous membrane filter was studied numerically and experimentally in the present work. We found that an electrical double layer around a single fiber expanded to a large extent at a low ion concentration, as in ultrapure water, and deformed toward the upstream of the fiber with increasing filtration velocity. Since an increase in the electrical double-layer thickness leads to a decrease in the electrical potential gradient, particles with the same polarity as the fiber approach the fiber more easily and are captured at a high filtration velocity. Experimental results also confirmed that the collection efficiency of polystyrene latex(PSL) particles through a PTFE filter became higher as the filtration velocity increased.