• Journal of KIISE:Computing Practices and Letters
• /
• v.8 no.3
• /
• pp.344-355
• /
• 2002

Having a variety of good characteristics against other pattern recognition techniques, Multilayer Perceptron (MLP) has been used in wide applications. But, it is known that Error Backpropagation (EBP) algorithm which MLP uses in learning has a defect that requires relatively long leaning time. Because learning data in pattern recognition contain abundant redundancies, in order to increase learning speed it is very effective to use online-based teaming methods, which update parameters of MLP pattern by pattern. Typical online EBP algorithm applies fixed learning rate for each update of parameters. Though a large amount of speedup with online EBP can be obtained by choosing an appropriate fixed rate, fixing the rate leads to the problem that the algorithm cannot respond effectively to different leaning phases as the phases change and the learning pattern areas vary. To solve this problem, this paper defines learning as three phases and proposes a Instant Learning by Varying Rate and Skipping (ILVRS) method to reflect only necessary patterns when learning phases change. The basic concept of ILVRS is as follows. To discriminate and use necessary patterns which change as learning proceeds, (1) ILVRS uses a variable learning rate which is an error calculated from each pattern and is suppressed within a proper range, and (2) ILVRS bypasses unnecessary patterns in loaming phases. In this paper, an experimentation is conducted for speaker verification as an application of pattern recognition, and the results are presented to verify the performance of ILVRS.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.39 no.2
• /
• pp.157-165
• /
• 2002

This paper presents the implementation of the speed control system for 3 phase induction motor using PD controller and neural networks. The PD controller is used to control the motor and to train neural networks at the first time. And neural networks are widely used as controllers because of a nonlinear mapping capability, we used feedforward neural networks(FNN) in order to simply design the speed control system of the 3 phase induction motor. Neural networks are tuned online using the speed reference, actual speed measured from an encoder and control input current to motor. PD controller and neural networks are applied to the speed control system for 3 phase induction motor, are compared with PI controller through computer simulation and experiment respectively. The results are illustrated that the output of the PD controller is decreased and feedforward neural networks act main controller, and the proposed hybrid controllers show better performance than the PI controller in abrupt load variation and the precise control is possible because the steady state error can be minimized by training neural networks.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2003.05a
• /
• pp.849-856
• /
• 2003

본 논문에서는 다층 퍼셉트론의 학습을 위한 새로운 두 단계 학습방법을 제안하였다. 첫 번째 단계는 국소최적해로 빨리 수렴하기 위해 Levenberg-Marquardt 알고리즘을 이용한 국소 탐색 단계이다. 두 번째 단계는 첫 번째 단계에서 찾은 국소최적해가 원하는 수준에 미치지 못할 경우 새로운 국소최적해로 벗어나기 위한 선형탐색을 기반의 터널링 단계이다. 이 방법은 연결가중치 공간에서 전역최적해를 빠르게 찾을 수 잇는 새로운 방법을 제공한다. 4가지 벤치마크 문제에 기존의 다층 퍼셉트론의 학습 알고리즘과 비교 실험을 통해, 제안된 알고리즘이 빠른 수렴 속도와 낮은 오차값을 가짐을 알 수 있었다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2002.10d
• /
• pp.568-570
• /
• 2002

최근, 분류기 쪽에서는 모듈라 학습을 이용한 방법들에 대해서 상당한 관심이 모아지고 있다. 모듈라 학습 방법은 divide and conquer 개념에 바탕을 두고 있기 때문에 복잡한 문제에 대해서 학습 질 측면이나 학습 속도 면에서 단일 분류기에 비해 좋은 결과들을 나타내고 있다. 인공신경망을 이용한 분류 방법 쪽에서도 이러한 연구들이 이루어지고 있다. 본 논문에서는 번호판 인식을 위한 간단한 형태의 모듈라 신경망을 제안하고 이의 성능을 평가하였다. 실험 결과, 일반적인 차량 번호판의 영상에서 성공적인 결과를 보였으며, 잡음에 의한 훼손된 번호판도 좋은 인식 결과를 보였다. 또한 인식률 측면 뿐만 아니라 학습 속도 면에서도 상당한 이득이 있었다.

• Journal of the Korean Chemical Society
• /
• v.53 no.1
• /
• pp.51-61
• /
• 2009

The purposes of this study are to analyze the learning contents on the measurement of reaction rate which is introduced in the high school ‘science’ and ‘chemistry II’ textbooks, and to revise the experiment appropriate to the learning contents. We examined 11 kinds of ‘science’ textbooks, 8 kinds of ‘chemistry II’ textbooks and 11 kinds of teacher’s manuals used in Korea and additionally surveyed teachers’ opinions on this subject. Most of textbook and teacher’s manuals described that ‘the reaction rate generally decreases through the time’, teachers’ conception also agreed with it. But most of experimental activities in the textbooks were inadequate to explain the concept that the reaction rate generally decreases with time. We analyzed the reasons and revised the experimental condition to solve this disagreement between the description in textbooks and an experimental result. Then we compared improved experimental result and theoretical prediction data. The improved experiment in this study is expected to help to describe the conception of chemical reaction rate in the textbook more clearly.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.39 no.3
• /
• pp.88-98
• /
• 2002

The multilayer perceptron (MLP) has several advantages against other pattern recognition methods, and is expected to be used as the learning and recognizing speakers of speaker verification system. But because of the low learning speed of the error backpropagation (EBP) algorithm that is used for the MLP learning, the MLP learning requires considerable time. Because the speaker verification system must provide verification services just after a speaker's enrollment, it is required to solve the problem. So, this paper tries to make short of time required to enroll speakers with the MLP based speaker verification system, using the method of improving the EBP learning speed and the method of reducing background speakers which adopts the cohort speakers method from the existing speaker verification.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.07a
• /
• pp.656-659
• /
• 2020

초해상화 딥러닝 기법은 학습 시 수렴하기까지 최소 수백 번의 에폭을 필요로 하며 오랜 시간이 걸린다. 최근, 영상 인식용 딥러닝 모델에서는 학습 수렴 속도를 향상시키기 위해 픽셀, 채널간 불필요한 상호연관 정보를 제거하는 Deconvolution 기술이 제안되었다. 본 논문에서는 최초로 Deconvolution 기술을 초해상화 딥러닝 방법에 적용하여 학습 수렴 속도 증가를 시도했다. 영상 인식 딥러닝 기법과 다르게 초해상화 딥러닝 기법은 이미지 특성 추출 부분과 이미지 복원 부분의 정보를 보존하는 것이 중요하기 때문에, EDSR을 Baseline 모델로 사용하여 양쪽 끝의 레이어는 기존의 Convolution 연산을 그대로 유지하고, 중간 레이어의 ResBlock 내의 Convolution 연산만 Deconvolution 연산으로 바꿔서 구성하였다. 초해상화 벤치마크 데이터셋을 사용한 실험 결과, 수렴속도가 빨라지지 않는 결과를 도출했다. 본 논문에서는 Deconvolution 기술이 Baseline 모델의 성능을 개선하지 못하는 이유를 초해상화 분야에서 기본적으로 적용되는 Residual Learning 기법 때문으로 분석했다.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.10 no.6
• /
• pp.597-604
• /
• 2000

본 연구에서는 전기자동차에 사용되는 전자식 차동 시스템의 신경망 모델을 제안한다. 차량이 곡선도로를 따라 주행할 경우 내측 바퀴와 외측 바퀴의 회전속도가 서로 달라야 진동이나 뒤틀림 없이 완만한 선회 주행을 할 수 있다. 전기자동차는 그 구조적 특성상 각각의 바퀴가 독립된 구동원을 갖는다. 이 때문에 일반 엔진 차량의 기어식 차동장치를 대신할 전자식 차동장치가 요구된다. 이러한 차동장치는 차량의 구조뿐만 아니라 차량의 주요 파라미터인 조향각 및 속도에 따라서 비선형적인 관계를 가지고 있어서 해석하기가 쉽지 않다. 따라서 이와 같은 비선형적인 관계 모델을 학습 능력을 가진 신경망에 의하여 모델링 함으로써 제어에 적용할 수 있다. 이를 실현하기 위해 제작한 전기자동차로 곡선도로를 주행하여 다양한 곡률과 주행속도에 따른 내측 외측 바퀴의 회전속도 데이터를 획득하고, 데이터의 비선형 특성을 고려한 차동 속도 제어기의 구조를 설계한다. 이 제어기에 적합한 모델은 신경망을 이용하여 실측 데이터를 학습시킴으로써 차동기능을 수행할 수 있는 제어기를 구현한다.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.18 no.2
• /
• pp.93-103
• /
• 2019

This study analyzed the impact of smartphone usage on walking speed during walking on two pedestrian walkways in Daejeon Metropolitan City. For the analysis, the video data about the actual use of smartphone was acquired and the walking speed was calculated based on the walking density of the pedestrian Level Of Service(LOS) presented in the Road Capacity Manual. Multiple regression analysis and decision tree using machine learning were used to analyze the impact of smartphone usage on walking speed, and as the explanatory variables, gender, disable smartphone, use of smartphone using auditory function, use of smartphone using visual function, LOS A, LOS B, LOS C were adopted. The result showed that LOS C had the highest impact on walking speed change and the women's group using their visual function was founded to have the slowest walking speed in LOS C. In particular, the author found that walking speed significantly decreased in the case of use of visual function rather than listening to music or the hearing on the phone.

• The Journal of the Acoustical Society of Korea
• /
• v.13 no.4
• /
• pp.60-68
• /
• 1994

In this paper we propose a new Hybrid learning rule applied to multilayer feedforward neural networks, which is constructed by combining Hebbian learning rule that is a good feature extractor and Back-Propagation(BP) learning rule that is an excellent classifier. Unlike the BP rule used in multi-layer perceptron(MLP), the proposed Hybrid learning rule is used for uptate of all connection weights except for output connection weigths becase the Hebbian learning in output layer does not guarantee learning convergence. To evaluate the performance, the proposed hybrid rule is applied to classifier problems in two dimensional space and shows better performance than the one applied only by the BP rule. In terms of learning speed the proposed rule converges faster than the conventional BP. For example, the learning of the proposed Hybrid can be done in 2/10 of the iterations that are required for BP, while the recognition rate of the proposed Hybrid is improved by about $0.778\%$ at the peak.