http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Differential Evolution with Adaptive Population Size for Target Localization in WSN
Andres Caceres Najarro,Iickho Song,Kiseon Kim 한국통신학회 2022 한국통신학회 학술대회논문집 Vol.2022 No.2
The population size (PS) plays a key role in the performance of any population-based evolutionary algorithm. In this paper, we propose two new techniques for adapting the PS, namely, parabolic and logistic reduction. In contrast to other PS adaptive techniques, the proposed techniques carefully reduce the PS at a higher rate. The proposed techniques together with the differential evolution (DE) are tested when solving the target node localization problem in terms of localization accuracy and computational complexity. Simulation results demonstrate that the DE with the proposed PS adaptive techniques provides better performance over the DE with other adaptive techniques, especially in terms of computational complexity.
Nonlinear Compensation Using Artificial Neural Network in Radio-over-Fiber System
Najarro, Andres Caceres,Kim, Sung-Man The Korea Institute of Information and Commucation 2018 Journal of information and communication convergen Vol.16 No.1
In radio-over-fiber (RoF) systems, nonlinear compensation is very important to meet the error vector magnitude (EVM) requirement of the mobile network standards. In this study, a nonlinear compensation technique based on an artificial neural network (ANN) is proposed for RoF systems. This technique is based on a backpropagation neural network (BPNN) with one hidden layer and three neuron units in this study. The BPNN obtains the inverse response of the system to compensate for nonlinearities. The EVM of the signal is measured by changing the number of neurons and the hidden layers in a RoF system modeled by a measured data. Based on our simulation results, it is concluded that one hidden layer and three neuron units are adequate for the RoF system. Our results showed that the EVMs were improved from 4.027% to 2.605% by using the proposed ANN compensator.
LSTM 기반 스태킹 앙상블 기법을 활용한 전력망 내의 전기자동차 충전 전력 수요량 예측 알고리즘
양창석(Changseok Yang),안드레스(Lismer Andres Caceres Najarro),김기선(Kiseon Kim) 한국통신학회 2021 한국통신학회 학술대회논문집 Vol.2021 No.6
최근 여러 기계학습 알고리즘과 딥러닝 신경망을 활용하여 전력망 내의 전기자동차 충전 전력 수요량을 예측하는 연구가 진행되고 있다. 하지만 전기자동차 사용자별 다양한 충전 행동 패턴과 시간대별로 다른 전기자동차 충전 요청 수요의 높은 변동성은 예측 성능을 저하하는 문제를 야기시킨다. 본 연구에서는 기계학습 알고리즘들과 장단기 메모리순환 신경망(Long Short-Term Memory)을 결합한 새로운 스태킹 앙상블 방법을 제안하였다. Caltech 대학 주차장의 EV 충전 사용 데이터 기반 모의 실험 결과, 제안한 방법이 기계학습 알고리즘과 순환 신경망 방법과 비교하여 RMSE과 MAE 두가지 성능 평가 지표에서 충전 전력 수요량 예측 오차를 개선함을 확인하였다.