网络函数

Methods The relationship of dynamic model′s input and output was analyzed and a basic structure of system identification model based on the principle of optimization in model reduction and the property of a function approximator to be composed of feed-forward networks was presented.

分析了控制系统动态模型的输入、输出关系，依据模型等效的最优化原理和神经网络任意逼近有限不连续函数的性质，提出神经网络辨识模型的基本结构。

The characteristics of B|spline basis function is first discussed, based on which B|spline neural network is proved to be a universal approximator.

1引言对于一个给定的非线性函数，我们都知道用神经网络可以以任意的精度来逼近它，原因就是神经网络是全局逼近器［1］。

The main reason for using neural network in time series analysis and system identification is that a trained neural network has the capability to approximate a arbitrary bounded continuous function.

在时间序列分析和系统辨识中，利用神经网络的一个主要原因就是前向神经网络逼近任意连续有界函数的能力。

At the same time,in order to insure the calculability of the objective function in the searching process of the particle swarm,neural network is used to model the voltage boosting circuit.

同时为了保证粒子群在搜索过程中目标函数的可计算，采用神经网络对升压电路进行建模，通过实验数据对神经网络进行训练并测试。

In order to solve the problems that generic ANN is lack of incremental learning and easily to make a wrong or a leak diagnosis, the paper has put forward a composite radial basis function network classifier oriented the diagnosis task, and studied the realization algorithm of this classifier in detail.

针对一般的神经网络不具备渐进学习功能、在故障诊断时易发生漏诊或误诊的问题，提出了一种面向诊断任务的组合径向基函数网络分类器，详细研究了该分类器的实现算法。

Fuzzy neural networks ; approximation capability ; continuous function ; density ; monolithic fuzzy neural networks

模糊神经网络；逼近能力；连续函数；稠密；单体模糊神经网络

Artificial neural networks has been widely used in pattern recognition, signal and information processing, adaptive controlment, and many other fields because of its self-organize ability, self-adapt ability, robustness and fault tolerance, and which are coming from its parallel processing structure and distribute information storage. But at the same time, they also bring some problems such as the difficulty of getting optimized topology structure and the great training time-consuming, which restrict its reality use.

人工神经网络的并行处理结构和分布式信息存储带来的自组织性、自适应性、鲁棒性和容错性，使得它们在实际应用中已表现出了强大的生命力，成为当今研究的热点之一，已经广泛的应用于模式识别、信号与信息处理、函数逼近、系统辨识和自适应控制的诸多领域，但同时带来了的最优拓扑结构难以确定，训练耗时巨大的问题，也限制了神经网络在实际中的应用。

In order to form wireless multihop networks with low energy,high connectivity,low interference and desirable route,the paper applied topology controlment scheme based on game theory and designed a novel utility function.

为了构建低耗能、高连通性、低干扰并具有合理路由的多跳无线网络，应用基于博弈论的网络拓扑结构控制解决方案，为节点设计了一种新颖的收益函数。

The convergence theorem of the neural networks algorithm and the theorem of numerical integration solution and its inferences are presented and proved.

摘要该文提出了一种基于三角基函数神经网络算法求解数值积分的新方法，提出并证明了神经网络算法的收敛定理和数值积分的求解定理及推论。

And further more it gives a proof that this optimization problem can be seen as a convex programming in exponential neural manifold.

即将神经网络看作为统计模型，使用常用的统计推断技术，通过优化学习代价函数，对网络参数进行调整，从而达到学习的目的。

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。