CAJ | 학술논문

文章阐述了一种改进共轭梯度理论神经网络优化算法。该方法是在传统共轭梯度算法（CG）基础上引入对输出权值进行优化的理念，故称其为输出权值优化共轭梯度算法（OWO - CG）。这种算法在进行学习时，首先根据误差函数利用共轭梯度法计算收敛因子，并修改输入层和隐含层的权值因子；接着，计算隐含层输出函数，利用相关输出权值优化理论构建并求解线性方程组得到输出层的权值因子；最后，计算误差函数，利用该算法不停地修正神经网络回路输出值与期望输出值之间的差值，直到满足精度要求为止。仿真验证结果表明，与传统共轭梯度算法相比，这种算法的学习过程更加迅速和准确。
문장천술료일충개진공액제도이론신경망락우화산법。해방법시재전통공액제도산법（CG）기출상인입대수출권치진행우화적이념，고칭기위수출권치우화공액제도산법（OWO - CG）。저충산법재진행학습시，수선근거오차함수이용공액제도법계산수렴인자，병수개수입층화은함층적권치인자；접착，계산은함층수출함수，이용상관수출권치우화이론구건병구해선성방정조득도수출층적권치인자；최후，계산오차함수，이용해산법불정지수정신경망락회로수출치여기망수출치지간적차치，직도만족정도요구위지。방진험증결과표명，여전통공액제도산법상비，저충산법적학습과정경가신속화준학。
An optimized algorithm of the neural network was elaborated in this paper based on the improved conjugate gradient theory. . The method was called the output weight optimized conjugate gradient algorithm(OWO - CG)be-cause the concept of optimized output weights was introduced into the method,with the traditional conjugate gradient algorithm(CG)as the basis. Using this method,several steps should be followed in learning. First,calculate the convergence factor by means of the error function and the conjugate gradient theory,and modify the weight factors of the input layer and the hidden layer. Then,calculate the output function of the hidden layer,and construct and solve the system of linear equations by using the relevant output weight optimization theory so as to gain the weight factor of the output layer. Finally,calculate the error function and constantly modify the difference between the output values and the expected values in the neural network circuit until the accuracy requirement is met. The simulation results showed that the learning process of this algorithm is faster and more accurate than that of the traditional conjugate gra-dient algorithm.