• 말소리와 음성과학
• /
• 제7권4호
• /
• pp.17-25
• /
• 2015

For blind source separation of convolutive mixtures, FDICA(Frequency Domain Independent Component Analysis) algorithms are generally used. Since FDICA algorithm such as Sawada FDICA, IVA(Independent Vector Analysis) works on the frequency bin basis with a natural gradient descent method, it takes much time to converge. In this paper, we propose a new method to improve convergence speed in FDICA algorithm. The proposed method reduces the number of iteration drastically in the process of natural gradient descent method by applying a weighted inner product constraint of unmixing matrix. Experimental results have shown that the proposed method achieved large improvement of convergence speed without degrading the separation performance of the baseline algorithms.

• ETRI Journal
• /
• 제28권1호
• /
• pp.59-66
• /
• 2006

Learning behaviors of a radial basis function network (RBFN) using a singular value decomposition (SVD) and stochastic gradient (SG) algorithm, together named RBF-SVD-SG, for odor sensing systems are analyzed, and a fast training method is proposed. RBF input data is from a conducting polymer sensor array. It is revealed in this paper that the SG algorithm for the fine-tuning of centers and widths still shows ill-behaving learning results when a sufficiently small convergence coefficient is not used. Since the tuning of centers in RBFN plays a dominant role in the performance of RBFN odor sensing systems, our analysis is focused on the center-gradient variance of the RBFN-SVD-SG algorithm. We found analytically that the steadystate weight fluctuation and large values of a convergence coefficient can lead to an increase in variance of the center-gradient estimate. Based on this analysis, we propose to use the least mean square algorithm instead of SVD in adjusting the weight for stable steady-state weight behavior. Experimental results of the proposed algorithm have shown faster learning speed and better classification performance.

• 한국정보처리학회논문지
• /
• 제4권11호
• /
• pp.2769-2779
• /
• 1997

본 논문에서는 공액기울기법과 터널링 시스템을 조합사용하여 신경망의 학습성능을 향상시킬 수 있는 효율적인 방법을 제안하였다. 빠른 수렴속도의 학습을 위하여 공액 기울기법에 기초한 후향전파 알고리즘을 사용하였고, 국소최적해를 만났을 때 이를 벗어난 다른 연결가중치의 설정을 위해 동적터널링 시스템에 기초한 후향전파 알고리즘을 조합한 학습 알고리즘을 적용하였다. 제안된 방법을 패리티 검사 및 패턴분류 문제에 각각 적용하여 기존의 기울기 하강법에 기초한 후향전파 알고리즘 및 기울기 하강법과 동적터널링 시스템을 조합한 후향전파 알고리즘방법의 결과와 비교 고찰하여 제안된 방법이 다른 방법들 보다 학습성능에서 우수함을 나타내었다．

• 전력전자학회:학술대회논문집
• /
• 전력전자학회 1998년도 Proceedings ICPE 98 1998 International Conference on Power Electronics
• /
• pp.293-297
• /
• 1998

This paper presents a newly developed speed sensorless drive using RLS based on Neural Network Training Algorithm. The proposed algorithm has just the time-varying learning rate, while the wellknown back-propagation algorithm based on gradient descent has a constant learning rate. The number of iterations required by the new algorithm to converge is less than that of the back-propagation algorithm. The theoretical analysis and experimental results to verify the effectiveness of the proposed control strategy are described.

• 전자공학회논문지B
• /
• 제32B권3호
• /
• pp.503-511
• /
• 1995

A batch-mode algorithm is proposed to increase the speed of learning in the error backpropagation algorithm with variable learning rate and variable momentum parameters in classification problems. The objective function is normalized with respect to the number of patterns and output nodes. Also the gradient of the objective function is normalized in updating the connection weights to increase the effect of its backpropagated error. The learning rate and momentum parameters are determined from a function of the gradient norm and the number of weights. The learning rate depends on the square rott of the gradient norm while the momentum parameters depend on the gradient norm. In the two typical classification problems, simulation results demonstrate the effectiveness of the proposed algorithm.

• 전력전자학회:학술대회논문집
• /
• 전력전자학회 1998년도 전력전자학술대회 논문집
• /
• pp.384-387
• /
• 1998

This paper presents a newly developed speed sensorless drive using RLS(Recursive Least Squares) based on Neural Network Training Algorithm. The proposed algorithm based on the RLS has just the time-varying learning rate, while the well-known back-propagation (or generalized delta rule) algorithm based on gradient descent has a constant learning rate. The number of iterations required by the new algorithm to converge is less than that of the back-propagation algorithm. The RLS based on NN is used to adjust the motor speed so that the neural model output follows the desired trajectory. This mechanism forces the estimated speed to follow precisely the actual motor speed. In this paper, a flux estimation strategy using filter concept is discussed. The theoretical analysis and experimental results to verify the effectiveness of the proposed analysis and the proposed control strategy are described.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 1993년도 한국자동제어학술회의논문집(국내학술편); Seoul National University, Seoul; 20-22 Oct. 1993
• /
• pp.1-7
• /
• 1993

In this paper we propose the new adaptive control algorithm by using S-G algorithm based on the error equations derived by Slotine. We verify the validity of the proposed controller and convergence of three type parameter estimation law based on S-G algorithm from the computer simulation.

• 한국정보과학회:학술대회논문집
• /
• 한국정보과학회 2004년도 가을 학술발표논문집 Vol.31 No.2 (2)
• /
• pp.697-699
• /
• 2004

Stochastic Neighbor Embedding(SNE) is a probabilistic method of mapping high-dimensional data space into a low-dimensional representation with preserving neighbor identities. Even though SNE shows several useful properties, the gradient-based naive SNE algorithm has a critical limitation that it is very slow to converge. To overcome this limitation, faster optimization methods should be considered by using trust region method we call this method fast TR SNE. Moreover, this paper presents a couple of useful optimization methods(i.e. conjugate gradient method and Newton's method) to embody fast SNE algorithm. We compared above three methods and conclude that TR-SNE is the best algorithm among them considering speed and stability. Finally, we show several visualizing experiments of TR-SNE to confirm its stability by experiments.

• 대한전자공학회논문지
• /
• 제20권6호
• /
• pp.11-17
• /
• 1983

본 연구는 explicit M.K.A.C. 알고리즘으로 마이크로프로세서를 사용하여 직류전동기 속도 제어를 하였다. 실험에 사용된 적응 제어 알고리즘으로는 먼저 지수함수적 가중 최소사승법 (exponentially wighted least square method: E.W.L.S) 알고리즘의 계산상의 불안정과 수행시간을 최소로 하기 위하여 시간 지연이 있는 E.W.L.S.알고리즘에 EDUt-_인수화법을 도입한 UDUt-인수화법 알고리즘을 사용하였고, 또한 gradient-type의 알고리즘으로도 초기에 dtatl-zollr을 갖는 직류전동기를 효과적으로 제어하기 위하여 gradient-type의 알고리즘에 smoothing polynomal과 상수ℓ을 사용한 SM gradient-쇼pe의 알고리즘을 제안하였다. UDTt인수화법 알고리즘을 적용한 결과 수행시간이 단축되었으며, SM gradient-type의 알고리즘의 경우는 dead-zone을 위한 기동전압을 사용하지 않고도 효과적인 속도제어를 할 수 있음을 확인하였다.

• 대한전기학회논문지:시스템및제어부문D
• /
• 제49권11호
• /
• pp.610-616
• /
• 2000

This paper presents an active noise control algorithm via Walsh transform domain controller learned by genetic algorithm. Typical active noise control algorithms such as the filtered-x lms algorithm are based on the gradient algorithm. Gradient algorithm have two major problems; local minima and eigenvalue ratio. To solve these problems, we propose a combined algorithm which consist of genetic learning algorithm and discrete Walsh transform called Walsh Transform Domain Genetic Algorithm(WTDGA). Analyses and computer simulations on the effect of Walsh transform to the genetic algorithm are performed. The results show that WTDGA increase convergence speed and reduce steady state errors.