电子与信息学报
電子與信息學報
전자여신식학보
JOURNAL OF ELECTRONICS & INFORMATION TECHNOLOGY
2013年
9期
2220-2226
,共7页
郭通%兰巨龙%李玉峰%江逸茗
郭通%蘭巨龍%李玉峰%江逸茗
곽통%란거룡%리옥봉%강일명
径向基函数神经网络%自适应粒子群优化%量子比特%流量预测
徑嚮基函數神經網絡%自適應粒子群優化%量子比特%流量預測
경향기함수신경망락%자괄응입자군우화%양자비특%류량예측
Radial Basis Function Neural Network (RBFNN)%Adaptive Particle Swarm Optimization (APSO)%Quantum bit%Traffic prediction
该文提出一种量子自适应粒子群优化算法,该算法中,粒子位置的编码采用量子比特实现,利用粒子飞行轨迹信息动态更新量子比特的状态,并引入量子非门实现变异操作以避免陷入局部最优。用该算法训练神经网络,实现了径向基函数(RBF)神经网络参数优化,建立了基于量子自适应粒子群优化RBF神经网络算法的网络流量预测模型。对真实网络流量的预测结果表明,该方法的收敛速度和预测精度均要优于传统RBF神经网络法、粒子群-RBF 神经网络法、混合粒子群-RBF 神经网络法和自适应粒子群-RBF 神经网络法,并且预测效果不易受时间尺度变化的影响。
該文提齣一種量子自適應粒子群優化算法,該算法中,粒子位置的編碼採用量子比特實現,利用粒子飛行軌跡信息動態更新量子比特的狀態,併引入量子非門實現變異操作以避免陷入跼部最優。用該算法訓練神經網絡,實現瞭徑嚮基函數(RBF)神經網絡參數優化,建立瞭基于量子自適應粒子群優化RBF神經網絡算法的網絡流量預測模型。對真實網絡流量的預測結果錶明,該方法的收斂速度和預測精度均要優于傳統RBF神經網絡法、粒子群-RBF 神經網絡法、混閤粒子群-RBF 神經網絡法和自適應粒子群-RBF 神經網絡法,併且預測效果不易受時間呎度變化的影響。
해문제출일충양자자괄응입자군우화산법,해산법중,입자위치적편마채용양자비특실현,이용입자비행궤적신식동태경신양자비특적상태,병인입양자비문실현변이조작이피면함입국부최우。용해산법훈련신경망락,실현료경향기함수(RBF)신경망락삼수우화,건립료기우양자자괄응입자군우화RBF신경망락산법적망락류량예측모형。대진실망락류량적예측결과표명,해방법적수렴속도화예측정도균요우우전통RBF신경망락법、입자군-RBF 신경망락법、혼합입자군-RBF 신경망락법화자괄응입자군-RBF 신경망락법,병차예측효과불역수시간척도변화적영향。
A novel Quantum Adaptive Particle Swarm Optimization (QAPSO) method is proposed. In this algorithm, the position encoding of the particle is achieved with quantum bits, and the state of quantum bit is updated dynamically with particle trajectory information. Then the mutation operation is performed by quantum non-gate to avoid falling into local optimum, which increases the diversity of particles. Afterwards, the Radial Basis Function (RBF) neural network is trained with QAPSO to implement the optimization of RBF neural network parameters. The network traffic prediction model is established based on the Quantum Adaptive Particle Swarm Optimization and RBF Neural Network (QAPSO-RBFNN). Forecasting results on real network traffic demonstrate that the convergence speed of the proposed method is faster and prediction accuracy is more accurate than that of traditional RBF neural network, the Particle Swarm Optimization and RBFNN (PSO-RBFNN), Hybrid Particle Swarm Optimization and RBFNN (HPSO-RBFNN), Adaptive Particle Swarm Optimization and RBF Neural Network (APSO-RBFNN). Furthermore, the forecasting effect of this method is stable on different scales.