• Proceedings of the Korean Information Science Society Conference
• /
• 2003.04c
• /
• pp.353-355
• /
• 2003

본 논문은 유전자 알고리즘의 다양성(diversity)을 유지하면서 동시에 수렴(convergence) 속도를 향상시키기 위한 새로운 선택기법을 제안한다. 이를 위해 적합도가 높은 염색체를 다음 세대로 전달하면서 동시에 적합도가 낮은 염색체에 대해서도 일정 수준 전달되게 하였다. 또한 기존의 설러 선택기법 중 가장 일반적으로 사용되는 토너먼트 선택 기법의 문제점을 고찰하고, 제안 알고리즘의 최적도 밀 수렴속도를 모의 실험을 통해 비교 및 분석한다. 실험 결과로부터 제안 알고리즘은 기존의 토너먼트 선택기법에 비해 우수함을 확인하였다.

• Proceedings of the Acoustical Society of Korea Conference
• /
• 1998.06e
• /
• pp.173-176
• /
• 1998

스테레오 음향 반향 제거기에서 발생하는 주요한 문제점들은 모노 환경에서와는 다르게 반향 경로 시스템의 긴 임펄스 응답으로 인한 느린 수렴속도와 원단화자 주위의 환경변화에 의한 최적해의 변화 등을 등 수 있다. 이러한 문제점들을 극복하기 위해 본 논문은 전송실에서의 환경 변화에 의한 반향 제거 성능저하와 저속의 수렴속도 및 과다한 계산량의 문제점을 해결하기 위하여 본 논문에서는 전송실의 환경 변화에 강인하고 계산량을 줄일 수 있는 Hyper-plane projection 알고리듬을 이용한 의사 스테레오 음향 반향 제거기를 제안한다.

• Communications for Statistical Applications and Methods
• /
• v.3 no.1
• /
• pp.39-49
• /
• 1996

비모수적 커널 회귀함수 추정법에서 평활량(bandwidth of smoothing parameter)의 선택은 아주 중요한 문제이다. 교차타당성(cross-validation) 방법에 의한 평활량은 최적평활량으로의 상대적 수렴속도(relative convergence rate)가 $n^{-1/10}$로 상당히 느리다는 것을 알고 있다. 본 연구는 삽입방법(plug-in method)에 의해 선택된 평활량의 상대적 수렴속도가 교차타당성 방법보다 더 빠른 $n^{-2/7}$이 됨을 보였다. 그리고 모의실험을 통하여 소 표본에서도 삽입방법이 교차타당성 방법보다 우수함을 입증하였다.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.45 no.2
• /
• pp.90-94
• /
• 2008

It is well-blown that the convergence rate gets worse when an input signal to an adaptive filter is correlated. In this paper we propose a new adaptive filtering algorithm that makes the convergence rate much improved even for highly correlated input signals. By introducing an orthogonal constraint between successive input signal vectors we overcome the slow convergence problem of the LMS algorithm with the correlated input signal. Simulation results show that the proposed algerian yields fast convergence speed and excellent tracking capability under both time-invariant and time-varying environments, while keeping both computation and implementation simple.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.22 no.11
• /
• pp.2451-2458
• /
• 1997

Adaptive channel equalization complished without resorting to a training sequence is known as blind equalization. In this paper, in order to increase the speed of the convergence and to reduce the steady-state mean squared error simulatneously, we propose the multi-stage DD(decision-direct) algorithm derived from the combination of the Sato algorithm and the decision-directed algorithm. In the starting stage, the multi-stage DD algorithm is identical to the Sato algorithm which guarantees the convergence of the equalizer. As the blind equalizer converges, the number of the level of the quantizers is increased gradally, so that the proposed algorithm operates identical to the decision-directed algorithm which leads to the low error power after the convergence. Therefore, the multi-stage DD algorithm obtains fast convergence rate and low steady state mean squared error.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.13 no.2
• /
• pp.130-135
• /
• 2002

A new Sign algorithm which has improved convergence speed is presented. The data-recycling technique, whose coefficients are multiply adapted in a symbol time period by recycling the received data, is applied to Sign algorithm which has few multiplications. Sign algorithm has very few multiplications and is the most easily implemented, but it gives small rate of convergence relative to others. The proposed algorithm combines the advatage of Sign algorithm, few multiplications, and the virtue of Data-Recycling LMS algorithm, simplicity and fast convergence. The results of computer simulation show that the proposed algorithm has 2 times faster convergence rate than that of LMS algorithm. Comparing to Data-Recycling LMS algorithm, in similar convergence conditions, it requires half fewer multiplications.

• The KIPS Transactions:PartB
• /
• v.8B no.2
• /
• pp.130-136
• /
• 2001

본 논문에서는 새로운 학습알고리즘의 비선형 주요성분분석 신경망을 이용한 영상데이터의 효율적인 특징추출에 대하여 제안한다. 제안된 학습알고리즘에서는 최적해로 수렴하는 과정에서 발생할 수도 있는 진동을 억제하여 빠른 속도의 수렴이 가능하도록 하기 위해 모멘트를 이용하였고, 국소최적해를 만났을 때 이를 벗어난 전역최적해로의 수렴을 위한 새로운 연결가중치의 설정을 위하여 동적터널링을 이용함으로써 빠른 수렴속도로 전역최적해에 수렴되도록 학습시킬 수 있다. 제안된 학습알고리즘을 이용한 신경망을 256$\times$256 픽셀의 간암영상과 128$\times$128 픽셀의 얼굴영상을 대상으로 실험한 결과, 기울기하강의 학습알고리즘을 이용한 기존 비선형 주요성분분석 신경망보다 우수한 수렴성능과 특징추출성능이 있음을 확인 할 수 있었다.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.19 no.5
• /
• pp.939-951
• /
• 1994

In adaptive equalizers based on least mean squares (LMS) algorithms, the convergence rate is determined by the convariance matrix of an input signal. When the eigenvalue spread of the convariance matrix is close to unity, the convergence rate is quite fast. In this paper, for fast convergence of LMS-based adaptive equalizers we propose a modified reference signal pertinent to the statistical channel. From the theoretical analysis and computer simulation, it is shown that the proposed modification method is quite effective for fast convergence of LMS-based adaptive equalizers.

• The KIPS Transactions:PartC
• /
• v.10C no.5
• /
• pp.617-624
• /
• 2003

This paper proposes a modified DCPC (Distributed Constrained Power Control, M-DCPC) algorithm that can improve the performance of a CDMA power control system. The control performance of the proposed method is verified using two performance measures : the SIR response of each mobile and the outage probability in a cell. As regards the SIR response, in simulations, the M-DCPC algorithm has shown a faster convergence and lower overshoot in transient time than the other power control algorithms when the desired SIR value was varying. For the outage probability. M-DCPC converged to a fixed outage rate faster than CSOPC while also maintaining the system capacity to make as high a connection as CSOPC. In particular, when the desired SIR was varying, CSOPC showed an abrupt outage probability increase during the desired SIR Increase, yet M-DCPC was unaffected.

• The Journal of the Acoustical Society of Korea
• /
• v.33 no.5
• /
• pp.335-340
• /
• 2014

In this paper, we propose a new sign subband adaptive filter to improve the convergence rate of the conventional sign subband adaptive filter which has been proposed to deal with colored input signal under the environment with impulsive noise. The existing sign subband adaptive filter does not increase the convergence speed by increasing the number of subband because each subband input signal is normalized by $l_2-norm$ of all of the subband input signals. We devised a new sign subband adaptive filter that normalizes each subband input signal with $l_2-norm$ of each subband input signal and increases the convergence rate by increasing the number of subband. We carried out a performance comparison of the proposed algorithm with the existing sign subband adaptive filter using a system identification model. It is shown that the proposed algorithm has faster convergence rate than the existing sign subband adaptive filter.