• 한국지능시스템학회:학술대회논문집
• /
• 한국퍼지및지능시스템학회 1997년도 춘계학술대회 학술발표 논문집
• /
• pp.185-188
• /
• 1997

In this paper, we execute identification, linearization, and control of a nonlinear system using recurrent neural networks. In general nonlinear control system become complex because of nonlinearity and uncertainty. And though we compose nonlinear control system based on the model, it is difficult to get good control ability. So we identify the nonlinear control system using the recurrent neural networks and execute feedback linearization of identified model, In this process we choose the optional linear system, and the system which will have to be feedback linearized if trained to follow the linearity between input and output of the system we choose. We the feedback linearized system by applying standard linear control strategy and simulation. And we evaluate the effectiveness by comparing the result which is linearized theoretically.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1993년도 하계학술대회 논문집 A
• /
• pp.224-226
• /
• 1993

Multi-layered neural networks offer an exciting alternative for modelling complex non-liner systems. This paper investigates the identification of continuous time nonliner system using neural networks with a single hidden layer. The digital low - pass filter are introduced to avoid direct approximation of system derivatives from sampled data. Using a pre-designed digital low pass filter, an approximated discrete-time estimation model is constructed easily. A continuous approximation liner model is first estimated from sampled input-out signals. Then the modeling error due to the nonlinearity is decreased by a compensator using neural network. Simulation results are given to demonstrate the effective of the proposed method.

• 센서학회지
• /
• 제30권5호
• /
• pp.290-294
• /
• 2021

Experts have designed popular and successful model architectures, which, however, were not the optimal option for different scenarios. Despite the remarkable performances achieved by deep neural networks, manually designed networks for classification tasks are the backbone of object detection. One major challenge is the ImageNet pre-training of the search space representation; moreover, the searched network incurs huge computational cost. Therefore, to overcome the obstacle of the pre-training process, we introduce a network adaptation technique using a pre-trained backbone model tested on ImageNet. The adaptation method can efficiently adapt the manually designed network on ImageNet to the new object-detection task. Neural architecture search (NAS) is adopted to adapt the architecture of the network. The adaptation is conducted on the MobileNetV2 network. The proposed NAS is tested using SSDLite detector. The results demonstrate increased performance compared to existing network architecture in terms of search cost, total number of adder arithmetics (Madds), and mean Average Precision(mAP). The total computational cost of the proposed NAS is much less than that of the State Of The Art (SOTA) NAS method.

• 한국산학기술학회논문지
• /
• 제10권6호
• /
• pp.1424-1432
• /
• 2009

충수돌기염 환자의 LoS(Length of Stay)를 예측하는 것은 병상의 운영에 적지 않은 영향을 준다. 본 논문에서는 Neural Networks와 Decision Tree를 이용하여 LoS와 연관이 높은 입력변수들을 찾아 그 의미를 분석하며, 찾아낸 입력변수들을 이용하여 다양한 LoS 예측 모형을 개발하고 그 성능을 비교하였다. 모형의 예측 정확성을 높이기 위하여 Bagging과 Boosting 등의 Ensemble 기법도 적용하였다. 실험 결과, Decision Tree 모형이 Neural Networks 모형보다 좀 더 적은 수의 속성을 가지고도 거의 통일한 예측력을 보였으며, Ensemble 기법 중에서는 Bagging 기법이 Boosting 기법보다 좋은 결과를 보여주었다. 의사결정나무 기법은 Neural Networks 기법에 비해 설명력이 있으며, 충수돌기염의 LoS 예측에 매우 효과적이었고, 중요 입력 변수의 선정에도 좋은 결과를 보여줌에 따라 향후 적극적인 기법의 도입이 필요하다고 할 수 있다.

• 한국지반공학회논문집
• /
• 제16권1호
• /
• pp.75-81
• /
• 2000

일반적으로 자연상태의 흙은 이방성을 나타내며, 이러한 흙의 이방성이 응력-변형률 거동에 미치는 영향은 매우 크다. 따라서 본 연구에서는 인공신경회로망 모델을 이용하여 압밀응력비 변화에 따른 정규압밀점토의 응력-변형률 거동을 모델링하고 비배수전단강도를 예측하여 보았다. 이때 사용된 신경회로망은 일반화된 델타규칙으로도 불리우는 오차역전파 학습 알고리즘을 이용한 다층신경회로망이다. 신경회로망의 학습은 인공퇴적 점토시료를 이용, 연직압밀응력과 압밀응력비를 다르게 정규압밀시킨후 비배수전단시험을 실시하여 얻어진 시험 결과를 이용하였고, 학습된 신경회로망을 이용하여 학습시 제외되었던 압밀응력비 상태에서의 비배수전단강도를 추론하여 본 결과 예측치와 실측치가 잘 일치하였다. 검토결과 실측치와 추론치 사이에는 결정계수($r^2$) 0.973 이상의 높은 상관관계가 있음을 확인하였다. 따라서, 본 연구결과는 점토의 비배수전단강도를 예측함에 있어서 인공신경회로망모델의 적용 가능성을 보여주었다.

• Communications for Statistical Applications and Methods
• /
• 제8권2호
• /
• pp.543-556
• /
• 2001

This study suggests integrated neural network modes for he stock price index forecasting using change-point detection. The basic concept of this proposed model is to obtain significant intervals occurred by change points, identify them as change-point groups, and reflect them in stock price index forecasting. The model is composed of three phases. The first phase is to detect successive structural changes in stock price index dataset. The second phase is to forecast change-point group with various data mining classifiers. The final phase is to forecast the stock price index with backpropagation neural networks. The proposed model is applied to the stock price index forecasting. This study then examines the predictability of integrated neural network models and compares the performance of data mining classifiers.

• 한국지능시스템학회논문지
• /
• 제2권3호
• /
• pp.50-57
• /
• 1992

Neural networks have become more popular recently and are now being applied to numerous fiedls. One of the major applications of neural networks is image recognition. Various image recognition system have been proposed so far, but there is no definite solution yet. In this paper, we propose a design of Fuzzy Neural Network Image Recognizer(FNNIR). Our model uses a fuzzy neural network model, named SONN[KIM90]. This model returns the information of the number of clusters and cluster and cluster center values for a given image data ste. Unlike the well-kinwn backpropagation technique, we do not need retraining for new data. Our newly designed image recongitionsystem FNNIR that uses fuzzy merger is proposed and experimented for a sample color image.

• 제어로봇시스템학회논문지
• /
• 제6권12호
• /
• pp.1070-1078
• /
• 2000

A nonlinear adaptive flight control system is proposed using a backstepping controller with neural network controller. The backstepping controller is used to stabilize all state variables simultaneously without the two-timescale assumption that separates the fast dynamics, involving the angular rates of the aircraft, from the slow dynamics which includes angle of attack, sideslip angle, and bank angle. It is assumed that the aerodynamic coefficients include uncertainty, and an adaptive controller based on neural networks is used to compensate for the effect of the aerodynamic modeling error. It is shown by the Lyapunov stability theorem that the tracking errors and the weights of neural networks exponentially converge to a compact set. Finally, nonlinear six-degree-of-freedom simulation results for an F-16 aircraft model are presented to demonstrate the effectiveness of the proposed control law.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2009년도 제40회 하계학술대회
• /
• pp.1112_1113
• /
• 2009

This paper proposes the method to predict the wind turbine power generation using logic-based fuzzy neural networks. To predict the wind turbine power generation neural networks, logic-based fuzzy neural networks, and fuzzy neural models have been considered. But the model considered in this paper can predict the wind turbine power generation with a less complex structure. The simulation results show the effectiveness of the proposed method.

• Smart Structures and Systems
• /
• 제5권5호
• /
• pp.507-515
• /
• 2009

The present work is an alternative methodology in order to balance a nonlinear highly flexible rotor by using neural networks. This procedure was developed aiming at improving the performance of classical balancing methods, which are developed in the context of linearity between acting forces and resulting displacements and are not well adapted to these situations. In this paper a fully experimental procedure using neural networks is implemented for dealing with the adaptive balancing of nonlinear rotors. The nonlinearity results from the large displacements measured due to the high flexibility of the foundation. A neural network based meta-model was developed to represent the system. The initialization of the learning procedure of the network is performed by using the influence coefficient method and the adaptive balancing strategy is prone to converge rapidly to a satisfactory solution. The methodology is tested successfully experimentally.