| 1. | In fact , it is a positive monotonically increasing function of the quotient of mav divided by the variance of weights
实际上， bam的抗噪声能力是mav和连接权方差之商的单增函数。 |
| 2. | The merits and limitations of genetic algorithms used in optimizing the connection weights of the neural network are discussed
摘要讨论了遗传算法优化神经网络连接权的优点及存在的局限性。 |
| 3. | The method of automatic fuzzy rules extraction based on fuzzy bp net researches hidden key attributes through deleting redundant linking weight
建立基于模糊bp网络的自动模糊规则提取方法，它通过删除冗余连接权的方法寻找到网络的隐含关键特征。 |
| 4. | In order to satisfy the requirement of the given precision , the connection power of the networks is studied and adjusted using the baekpropagation training algorithm ( bp algorithm )
采用误差反向传播算法( bp算法)对网络的连接权值进行学习和调整，以满足给定的精度要求。 |
| 5. | It should be mentioned that all the results obtained in this chapter is relevant to the hypnosis that the weight matrix is symmetric . chapter 5 is made up of two sections
而这些结果都不是以连接权矩阵具有对称性作为前提，所以部分结果涵盖了原有的在对称连接权矩阵条件下的前人的一些结果 |
| 6. | The algorithms is carried into training connection weights of nn and simulation experiments show the arithmetic can escape local optima and improve learning speed of nn to some extent
将其用于调整神经网络的连接权值，实验证明该方法可克服神经网络训练的局部最优解问题，并在一定程度上提高神经网络的学习速度。 |
| 7. | This algorithm uses the quotient as the fitness of each individual and employs pseudo - relaxation method to adjust individual solution when it does not satisfy constraining condition any more after genetic operation
这种方法应用遗传算法和准松弛方法来得到bam的可行解，以mav和连接权方差之商为个体适应度函数，并应用准松弛方法来调整不满足约束条件的个体。 |
| 8. | Anew neural network algorithm for decision level fusion is also presented in this dissertation . the architecture of this network is novel . it is the thresholds , not the conjunction weights , which are modified , when the network is being trained
本文还提出了一种新的神经网络算法用于决策层融合目标识别，该网络结构新颖，网络训练时修改的是门限而不是连接权值。 |
| 9. | Ann has strong parallel running , fault - tolerant and self - learning " capacity . ann can finish the gain of knowledge by the model sampling and . memory the knowledge into the weights of topological structure
神经网络具有很强的并行性、容错性和自学习能力，通过对典型样本的学习，完成知识的获取，并将知识分布存储在神经网络的拓扑结构连接权值中，用来对未知样本进行识别。 |
