网络函数

A direct correspondence is shown between MTS fuzzy model with the determining consequence structures and the radial basis function networks,.

该方法表明具有确定后件结构的 MTS模糊模型与径向基函数网络之间有一种直接对应关系，基于这种对应，我们可把 MTS模型的前件结构确定和后件结构辨识分开，利用径向基函数网络的学习特性和其它学习算法相结合来得到模糊模型。

Therefore an artificial nervenetwork with better effect of imminence should be designed so that it can be used in the field of technological innovation and achieve good results.

本文利用径向基函数神经网络这一技术手段，提出了基于径向基函数神经网络的技术创新风险评价模型，为技术创新项目的风险评价提出了一种新颖、高效的方法。

Optimal algorithm of combined nonlinear Hopfield network appears powerful validity in solving nonlinear planning, which includes nonlinear objective function, linear constrains and high-demension of decision-making variants, for its ability of nonlinear parallel computation. It is prominent among optimal algorithms because of its function of simply implement with electrocircuit hardware. Genetic algorithm is expressly suitable for optimal calculation regarding the planning of massive, highly nonlinear, inconsecutively differentiable and multiobjective function as well as objective function without analytical expression. However, it inclines to prematurity, as well as its limitation in ability of partial optimal search. Introducing the optimal algorithm of niche genetic simulated annealing to standard Genetic algorithm, can therefore improve the full-scale or partial search ability of Genetic algorithm effectively. It has a far-flung perspective in the field of systemic planning of water pollution control.

组合式非线性Hopfield网络优化算法所具有的非线性大规模并行计算能力在求解具有非线性目标函数、线性约束条件及高维决策变量的非线性规划问题方面显示出了强大的生命力，它易于电路硬件实现的功能更是在优化算法中独树一帜；遗传算法采用概率搜索技术，不受目标函数与约束条件的限制，特别适合大规模、高度非线性的不连续可微的多峰目标函数及无解析表达式的目标函数的规划问题的优化计算，但其存在容易早熟、局部寻优能力较差等缺点，本文在标准遗传算法中引入小生境技术及模拟退火算法有效地改善遗传算法的全局和局部搜索性能，提高了全局最优解的寻优质量，小生境遗传退火模拟优化算法在水污染控制系统规划中的应用前景极为广阔。

The results show that small world network model could explain the big change of intensiveness and connectedness in forming process. And individual selection value function could explain the stable intensiveness, connectedness in growing and mature process. Combined the both, the characteristics of the whole process could be reflected.

研究表明，小世界网络模型模拟的集群网络具有集群形成期网络整体密集性和连通性剧烈变化的特征，而个体选择价值函数模型可以解释集群网络的成长期、成熟期的密集性和连通性趋于稳定的特征，将2种演化规则相结合，则可以反映集群网络从形成期到成熟期的各生命周期特征。

Using RBF neural network substitutes BP network as the approximation of implicit performance function, and the response surface is formed in the sphere of hyper-pyramid at iteration step.

采用径向基函数神经网络，代替目前常用的BP网络，迭代过程中在超锥体的范围内构造响应面，逼近隐式的非线性功能函数。

In this paper continuous differentiable conditions of output response functions of bidirectional associate memory neural networks are reduced to Lipschitz condition.

因此，本文将双向联想记忆神经网络的输出响应函数连续可微的假设削弱为满足Lipschitz条件，通过引入Lyapunov函数，利用不等式的方法，证明了双向联想记忆神经网络全局指数稳定性的一个定理。

For the data set with noises, a regularization intropolation method is proposed according to regularization theory. The relation between the regularization intropolation method and radial basis function method is analysed and structure of regularization neural networks is proposed. RBF neural network is introduced by mortifying the regularization neural networks. Finally the approximation capacity of RBF neural networks is analysed. 4. A method of selecting the centers of hidden layer neurons of RBF neural networks is proposed.

首先从精确内插问题开始对RBF神经网络进行讨论，然后根据正则化理论提出了在数据集带有噪声的情况下的内插方法，并分析了这种内插方法和径向基函数方法之间的密切联系以及其对应的正则化神经网络结构，其次对正则化神经网络进行了修改，得到正则化神经网络的简化形式—RBF神经网络，最后分析了RBF神经网络的逼近性能。

The penalty optimal brain surgeon is a post-training algorithm and it has extreme high complexity. The OBS oriented compute model implemented neural network training and OBS pruning simultaneously, by taking the OBS pruning case as a penalty term of neural network objective functions based on optimized structure of the regularization method included in the neural network training process. It both maintained the OBS's accuracy and had regularization method's high efficiency. Raised the generalization of neural network model.

最优脑外科过程是一种训练后网络剪枝算法，计算的复杂度非常高，通过把剪枝条件以惩罚项的形式纳入神经网络的训练目标函数中，把正则化方法的结构优化蕴涵于网络训练过程，构建面向最优脑外科过程的计算模型，实现网络训练过程和最优脑外科过程并行剪枝，既保持了最优脑外科过程的准确性，又具有正则化的高效性，提高了神经网络模型的泛化性能。

Secondly, contraposes the limitation of the traditional harmonic current detecting method based on neural network, presents the method that combines the neural network and auto adaptive harmonic current detecting method based on the principle of noise each other eliminating.As the characteristics that are small calculative quantity, fast converge, without local minimal point of Redial Basis Function, this dissertation forms a new harmonic current detecting method based on RBF.

其次，针对传统的谐波电流检测方法的缺陷，提出将神经网络与基于噪声抵消原理的自适应谐波检测相结合，利用径向基函数运算量小、收敛快、无局部极小值等优点，构造了一种基于径向基函数神经网络的谐波电流检测方法，仿真结果表明该检测方法具有很好的动态响应及畸变电流检测精度。

Conventionally the electric load forecasting can hardly attain a result whose accuracy meets what's required. A short-term load forecasting model is therefore developed to solve the problem, based on the process neural network of which the input is the function of time and the high forecasting accuracy is available. Describes the structure of the model, discrete data fitting method by the expansion of function orthogonal basis and learning algorithm.

针对目前常用方法在解决负荷预测问题时，结果往往难以达到工程要求精度的现状，利用过程神经网络输入为时间函数以及预测精度高的特点，建立了基于过程神经网络的电力系统短期负荷预测模型；给出了模型的结构，基于函数正交基展开的离散数据拟合方法以及模型的学习算法。

The split between the two groups can hardly be papered over.

这两个团体间的分歧难以掩饰。

This approach not only encourages a greater number of responses, but minimizes the likelihood of stale groupthink.

这种做法不仅鼓励了更多的反应，而且减少跟风的可能性。