oja

In this paper, a fast learning algorithm was proposed on the Oja's rule of unsupervised learning principle and delta rule of supervised learing principle The synaptic weights of the networks is initiated with them.

本文提出一种基于非监督学习的Oja规则和有监督学习的delta规则的快速学习算法.用改进后的结构和学习方法初始化前馈神经网络的权值；

A New ANN Harmonic Analysis Algorithm Based on Oja's Learning Rule

基于Oja学习规则的ANN谐波分析算法

While using continuous time neural network described by the E. Oja. learning rule ( Oja-N) for computing real symmetrical matrix eigenvalues and eigenvectors, the initial vector must be on Rn unit hyper-sphere surface, otherwise, the network may produce limit-time overflow.

当把Oja学习规则描述的连续型全反馈神经网络（Oja-N）用于求解矩阵特征值特征向量时，网络初始向量需位于单位超球面上，这给应用带来不便。

But in some cases, the convergent property of the algorithm is not satisfactory. A new ANN algorithm for analyzing harmonics in power system based on ADALINE model and Oja's learning rule is described.

在ADALINE模型的基础上，提出了一种新的、适用于电力系统谐波分析的ANN算法，该算法的训练过程采用了Oja学习规则，较好地保证了算法的收敛性。

Oja's idea of statistical principal components analysis in recursion network and S.Kung's adaptive principal components extraction ( APEX) algorithm in feature extraction are pointed out.

Oja将统计主元分析思想用于递推网络与S。

Using the normalized version of orthogonal Oja method tracks the signal subspace fast and adaptively.

最后，应用归一化正交Oja快速自适应地追踪信号子空间。