• The Journal of the Korea Contents Association
• /
• v.3 no.3
• /
• pp.98-102
• /
• 2003

For elucidating the key role of hidden neurons in information processing of Multilayer perceptrons(MLPs), we prove that the correlation coefficient between weighted sums to hidden neurons decreases under element-wise nonlinear transformations. This is verified through training of MLPs for an isolated word recognition problem. From this result, we can say that the element-wise nonlinear functions reduces redundancy in the information contents of hidden neurons.

• Proceedings of the Korea Inteligent Information System Society Conference
• /
• 2005.05a
• /
• pp.169-185
• /
• 2005

신경계의 뉴런 구조는 흥분 뉴런과 억제 뉴런으로 구성되며 각각의 흥분 뉴런과 억제 뉴런은 주동근 뉴런(agonistic neuron)에 의해 활성화되며 길항근 뉴런(antagonist neuron)에 의해 비활성화 된다. 본 논문에서는 인간 신경계의 생리학적 뉴런 구조를 분석하여 퍼지 논리를 이용한 생리학적 퍼지 신경망을 제안한다. 제안된 구조는 주동근 뉴런에 의해 흥분 뉴런이 될 수 있는 뉴런들을 선택하여 흥분시켜 출력층으로 전달하고 나머지 뉴런들을 억제시켜 출력층에 전달시키지 않는다. 신경계를 기반으로 한 제안된 생리학적 퍼지 신경망의 학습구조는 입력층, 학습 데이터의 특징을 분류하는 중간층, 그리고 출력층으로 구성된다. 제안된 퍼지 신경망의 학습 및 인식 성능을 평가하기 위해 정확성이 요구되는 의학의 한 분야인 기관지 편평암 영상인식과 영상 인식의 주요 응용 분야인 차량 번호판 인식에 적용하여 기존의 신경망과 성능을 비교 분석하였다. 실험 결과에서는 제안된 생리학적 퍼지 신경망이 기존의 신경망보다 학습 시간과 수렴성이 개선되었을 뿐만 아니라, 인식에 있어서도 우수한 성능이 있음을 확인하였다.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2008.06a
• /
• pp.149-154
• /
• 2008

FCM 기반 RBF 네트워크는 서로 다른 학습 구조가 결합된 혼합형 모델로서, 입력층과 중간층의 학습 구조는 FCM 알고리즘을 적용하고, 중간층과 출력층 사이의 학습 구조는 Max_Min 신경망을 적용한다. 입력층과 중간층의 학습시 입력벡터와 중간층의 노드중에서 중심과 입력벡터간의 가장 가까운 노드를 승자 노드로 선택하여 출력층으로 전달한다. 그리고 중간층과 출력층 사이의 학습 구조는 Max_Min 신경망을 적용하여 중간층의 승자 뉴런이 출력층의 입력벡터로 적용한다. 하지만 많은 패턴이 입력벡터로 제시될 경우 학습 성능이 저하되는 단점이 있다. 따라서 본 논문에서는 중간층과 출력층의 학습 구조인 Max_Min 알고리즘의 학습 성능을 개선시키기 위해 퍼지 제어시스템을 이용하여 학습률을 동적으로 조정하는 퍼지 제어 기법을 이용한 FCM 기반 RBF 네트워크를 제안한다. 제안된 방법의 학습 성능을 평가하기 위하여 컨테이너 영상에서 추출한 숫자, 영문 식별자를 학습 데이터로 적용한 결과, 기존의 ART2 기반 RBF 네트워크보다 학습 시간이 적게 소요되고, 학습의 수렴성이 개선된 것을 확인하였다.

• Journal of Intelligence and Information Systems
• /
• v.11 no.2
• /
• pp.81-103
• /
• 2005

The Neuron structure in a nervous system consists of inhibitory neurons and excitory neurons. Both neurons are activated by agonistic neurons and inactivated by antagonist neurons. In this paper, we proposed a physiological fuzzy neural network by analyzing the physiological neuron structure in the nervous system. The proposed structure selectively activates the neurons which go through a state of excitement caused by agonistic neurons and also transmit the signal of these neurons to the output layers. The proposed physiological fuzzy neural networks based on the nervous system consists of a input player, and the hidden layer which classifies features of learning data, and output layer. The proposed fuzzy neural network is applied to recognize bronchial squamous cell carcinoma images and car plate images. The result of the experiments shows that the learning time, the convergence, and the recognition rate of the proposed physiological fuzzy neural networks outperform the conventional neural networks.

• The Journal of the Korea Contents Association
• /
• v.4 no.3
• /
• pp.99-105
• /
• 2004

This paper proposes optimal learning rates in the gradient descent training of multilayer perceptrons, which are a separate learning rate for weights associated with each neuron and a separate one for assigning virtual hidden targets associated with each training pattern Effectiveness of the proposed error function was demonstrated for a handwritten digit recognition and an isolated-word recognition tasks and very fast learning convergence was obtained.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.8 no.5
• /
• pp.29-38
• /
• 1998

This paper proposes the neural network self-tuning PID controller for the load frequency control of 2- areas power system, namely, the prompt convergence of frequency and tie-line power flow deviation. The neural network applied to computer simulation consists of neurons of two inputs, ten hiddens and tliree outputs layer. Neurons of two inputs layer receive the error and its change rate of the system and cutputs layer consists of three neurons for the parameters of the PID controller. The simulation results shows that the proposed neural network self-tuning PID controller is superior to conventional control t~:chniques(Optimal, PID) in dynamic response and control performance.

• Journal of the Korea Society of Computer and Information
• /
• v.19 no.9
• /
• pp.33-38
• /
• 2014

The FCM based hybrid RBF network is a heterogeneous learning network model that applies FCM algorithm between input and middle layer and applies Max_Min algorithm between middle layer and output. The Max-Min neural network uses winner nodes of the middle layer as input but shows inefficient learning in performance when the input vector consists of too many patterns. To overcome this problem, we propose a dynamic learning rate control based on fuzzy logic. The proposed method first classifies accurate/inaccurate class with respect to the difference between target value and output value with threshold and then fuzzy membership function and fuzzy decision logic is designed to control the learning rate dynamically. We apply this proposed RBF network to the character recognition problem and the efficacy of the proposed method is verified in the experiment.

• The Journal of the Acoustical Society of Korea
• /
• v.13 no.6
• /
• pp.5-14
• /
• 1994

This paper is about an experiment of speaker-independent automatic Korean spoken words recognition using Multi-Layered Perceptron and Error Back-propagation algorithm. The object words are 50 citynames of D.D.D local numbers. 43 of those are 2 syllables and the rest 7 are 3 syllables. The words were not segmented into syllables or phonemes, and some feature components extracted from the words in equal gap were applied to the neural network. That led independent result on the speech duration, and the PARCOR coefficients calculated from the frames using linear predictive analysis were employed as feature components. This paper tried to find out the optimum conditions through 4 differerent experiments which are comparison between total and pre-classified training, dependency of recognition rate on the number of frames and PAROCR order, recognition change due to the number of neurons in the hidden layer, and the comparison of the output pattern composition method of output neurons. As a result, the recognition rate of $89.6\%$ is obtaimed through the research.