提出了一种综合考虑了训练误差和检验误差的评价神经模糊模型性能的误差性能指标。
A performance index of error was presented. This index is a kind of evaluation of neural fuzzy model performance and synthetically considered training error and checking error of NFS.
给出了应用实例。还研究了混沌参数与训练误差的关系，提出了混沌参数的调整步骤及应用。
And it studied relation between chaos mapping parameters and training error, puts forward chaos mapping parameter adjust steps and applications.
该技术首先对神经网络集成中的个体之间进行负相关处理提高个体的差异度，然后选择训练误差较小的个体来提高个体的精确度。
During the process of training, individual networks are trained using negative correlation learning to improve their diversity, and then networks with small errors are chosen to improve the accuracy.
又将蚂蚁行经路径上的存储单元存放的参数值赋予BP网络训练，而存储单元存放的参数和训练误差值亦随BP网络训练误差的调整而改变。
Parameters value of sites in ants' route was bestowed on BP network, while parameters and training error stored in stored units were changed along with the adjustment of training error of BP network.
该模式采用直接预报的方法，避免了误差累积效应；对训练集实时更新，以保证模式的不断更新；模式的输入比较简单，便于应用。
This model adopted direct prediction method, so error accumulation effect was avoided. The training data set was real-time updated so that the model was always the newest.
对目标误差、网络的学习率和训练次数进行了具体的优化。
The optimum programs are made for the target error, the learning of Neural Network, the time of training.
结果表明：用L -M规则进行神经网络学习训练可使网络收敛快，误差小。
The network lessens quicker and the error less trained with the L-M rule .
若这些参数超出了范围，训练装置将发出警报，以便提醒学员纠正误差。
If these parameters go beyond the ranges, the trainer will alarm to warn the learner to correct the errors.
他们密切监视克隆工程，防止任何生化误差，通过彻底的训练将任性的克隆人拉回整齐划一的队伍。
They closely monitor their cloning projects for any deviations in biochemistry, subjecting wayward clones to extensive conditioning to pull them back in line.
证明了在存在噪声干扰及网络设计误差的情况下，网络训练过程具有指数收敛性和鲁棒平稳性。
The exponential convergence of the training process and its robust stability to the noise perturbances and the network design errors are also proved.
用训练样本集对网络训练后，检验样本的预测结果与实际值最大误差为0.97%。
The model was trained with training sample aggregation. The maximum error between the forecasted and real value was 0.97%.
利用误差反向传播的改进算法对样本数据进行训练，并用另外的一些样本数据验证模型的应用效果。
Using the improved error backward propagation, the model is trained with stylebook data and validated its effect by other stylebook data.
在满足同样精度情况下，选用RBF神经网络建模，训练速度很快，基本误差小，较好地解决了风险补偿费率的估计问题。
Compare with other neural networks, the RBF neural network have the features of training quickly and little errors in estimating risk compensation rate.
探讨了采用神经网络校准仪器非线性的方法，并用递推预报误差算法训练神经网络。
The method of calibrating the nonlinearity of the instrument by using nervous network is investigated. The nervous network is trained by recurrence forecast error algorithm.
本文介绍了动态对角递归网络，并针对BP算法收敛慢的缺点，将递推预报误差学习算法应用到神经网络权值和域值的训练。
To overcome the slow convergence of the BP algorithm, recursive prediction error algorithm is proposed, which can train both the weight and the bias.
由于粗神经网络的误差传递函数不可微，所以采用遗传算法来训练粗神经网络。
Because the error transfer function of rough neural network is not differentiable, genetic algorithms are applied for training the network.
构造了局部连接的前馈神经网络的结构和基于递推预报误差的网络训练算法；
The structure of a partially connected feed forward neural network and the training algorithm based on the recursive prediction error are constructed.
其次，从误差反传算法在预测中存在的问题入手，提出一种混合训练算法。
Secondly, the hybrid training algorithm is proposed on the base of the Error Back-propagation algorithm's disadvantage analysis in the paper.
它通过构造伪输出辨识被控对象参数，引进反馈误差，实现对解耦神经网络的在线训练。
It can identify the parameters of a controlled object by forming a fake output and bring in a feedback error for performing an on-line training to decouple the neural network.
应用误差逆传播（BP）神经网络方法，对高含沙流体极限剪应力进行了多次训练。
Data of ultimate shear stress of hyper concentration flow are trained several times by Back Propagation (BP) neural network method.
采用信号四阶和六阶统计量提取信号特征，使用新设计的误差函数训练RBF神经网络，使得识别的效率和正确度得到了明显的改善。
The forth-order and sixth-order cumulants of received signal are adopted for features extraction while RBF neural networks with a new designed training cost function being used for classifier.
以同样方法对复合氨基酸注射液进行测定，通过训练好的网络进行色氨酸、酪氨酸含量的计算，相对误差分别为4.0％和2.6％。
The method has been applied to the determination of tryptophan and tyrosine in compound amino acid injection, and the relative errors were 4.0% and 2.6% .
根据国家标准试验火数据进行网络训练，系统误差小于试验火标准误差要求，表明了算法的有效性和可行性。
The network is trained using country testing fire data, system error is less than standard error, which shows the validity and feasibility of the algorithm.
该算法利用预测误差阈值进行样本的取舍，在尽量保留有用信息的情况下减小样本训练规模。
This algorithm USES the prediction error threshold to retain the useful information to decrease sample training scale.
应用神经网络的误差反向传播算法(BP)和大量的实测数据样本训练出了能在线诊断四种加工状态的BP模型并成功地诊断了实际加工状态。
The BP algorithm of Artificial Neural Networks and lots of experimental samples were used in training the BP model which succeeded in diagnosing four kinds of operational status.
许多实践者使用缺省损失函数(如，均方误差)训练和挑选最好的模型。
Many practitioners train and pick the best model using the default loss function (e. g., squared error).
采用了收敛速度较快的递推预报误差算法训练神经网络。
A recursive prediction error algorithm which converges fast is applied to tra.
此外，提出用带输出误差死区的混合BP算法训练神经元分类器，提高了网络学习训练速度和分类准确性。
Further, a hybrid BP algorithm with dead interval of error is derived for training the neural classifier in order to increase training speed and classification accuracy.
实例计算表明，与BP算法及BP与GA结合算法比较，该方法在提高误差精度的同时可以加快训练收敛的速度。
Simulation results of practical example show that the method can improve the calculation accuracy and the speed of the convergence process compared with BP and BP trained by GA.
实例计算表明，与BP算法及BP与GA结合算法比较，该方法在提高误差精度的同时可以加快训练收敛的速度。
Simulation results of practical example show that the method can improve the calculation accuracy and the speed of the convergence process compared with BP and BP trained by GA.
应用推荐