训练数据

一种基于NA假设的训练数据自动构造方法

Training Data Acquisition Based on NA Assumption

由于训练数据中缺乏对链接的标记，但预测时却需要找出用户感兴趣的链接，这就使得Web目录页面链接推荐问题相当困难。

Therefore , the problem of link recommendation in Web index page is quite difficult since the training data lacks links ' label while prediction for links of interest in a new Web index page is required .

对基于训练数据的最小均方（LMS）算法进行了研究；

The paper makes some researches on training symbol-assisted least-mean-square ( LMS ) algorithm .

其次，在将BP网络用于色彩匹配的过程中，设计了合理的网络结构，并对训练数据进行了规格化处理。

Then , in the course of applying BP network to color matching method , designing the reasonable network structure and standardizing the data are discussed .

基于SVG技术的试验训练数据信息可视化方法研究

Method of Information Visualization Based on the SVG Technique for the Data of Testing and Training

对多层ANN的结构和向后传播算法进行了设计，提出了移动窗口和事件子视图等概念，通过提取审计事件类型的方法，采样了ANN的训练数据和测试数据。

Structure and back propagation arithmetic of multilayer ANN is designed . Concept of moving window and event sub-view etc. is presented .

利用训练数据对LVQ和改进算法的BP神经网络进行训练后，利用测试数据对模型的检测效果进行验证。

Using the simulation data trains LVQ network and ameliorated BP network , then detection effect is validated by actual data .

用BP神经网络训练数据，调整权值和故障输出，建立刀具故障在线监测小波神经网络模型。

Those data is exercised BP NN , and adjust the value and output of faults . Finally a model of wavelet NN to inspect on-line tools faults is established .

基于模糊矢量量化以及隐马尔可夫模型（FVQ/HMM）的说话人识别研究：FVQ/HMM作为HMM的特殊形式，其模型参数数量较传统HMM少，模型学习对训练数据量要求不高；

Study of speaker recognition based on FVQ ( Fuzzy VQ ) / HMM : It is the special form of HMM.

在ClassificationWorkbench中选择包含训练数据文档文本的字段的地方，必须同样使用这个名称。

This name must be used again in the Classification Workbench , where you have to choose a field that contains the document text of your training data .

首先对实际生产中的原始数据，经过过失误差剔除及滤波处理后得到一套训练数据和校验数据组成训练样本，然后采用BP神经网络进行训练，得到组分纯度的非参数模型。

First by filter process of practical primal data , the training samples are gained . Then the nonparametric model on component purity is developed by use of BP neural network .

分别采用二次核函数以及高斯RBF核函数，利用训练数据对线性和非线性系统进行黑箱辨识。

According to the train data , linear and nonlinear systems ' black_box identification is performed by using SVM with quadric polynomial and Gaussian RBF kernel respectively .

针对语音识别中由于强噪声的影响而引起的Lombard和Loud效应进行研究，提出了基于训练数据的加性噪声和Lombard及Loud效应的联合补偿法。

This paper proposes a unified approach for the noisy Lombard and Loud speech recognition based on training data compensation .

最后，本文介绍了基因算法，并给出用基因算法（GeneticAlgorithm）训练数据的详细实验流程以及得到的实验数据。

In the end , Genetic Algorithm is brought out and introduced comprehensively . Moreover . the paper puts emphasis on the experimental flow that Genetic Algorithm is used to train data in detail and the results which are supplied in the laboratory .

阐述了在Excel平台上建立的BP网络模型水质评价系统的方法和要点，以及网络模型结构的设计、训练数据的处理、网络的训练和评价结果的仿真情况。

To establish BP network model water quality evaluation system based on Excel platform . The design of network model structure , treat of train data and simulation of evaluation result were discussed .

首先将整个训练数据集分为多个小的子集，然后同时运行多个CPU处理器处理每一个分离的数据集。

Specifically , the parallel SMO first partitions the entire training data set into smaller subsets and then simultaneously runs multiple CPU processors to deal with each of the partitioned data sets .

本文分别针对数据的预处理、训练数据集大小以及输入向量的大小分别进行了研究，以确定使用BP神经网络预测的一个最佳参数组合。

Raw data preprocessing , the size of training data , as well as the size of input vectors have been studied separately , to find out the best parameter set of BP neural network prediction .

其次，研究了一种利于计算机建模的市场需求组合预测方法，即采用Logit回归法训练数据的市场需求组合预测模型。

Secondly , a computationally - convenient means of combined forecasts is studied , which using Logit regression to train set data .

传统的度量方法，例如误差矩阵和Kappa系数，直接在训练数据上估计分类模型的性能。

In the classical fashions , e.g. error matrix and Kappa coefficient , the performance of the classification models is estimated directly on the training data .

针对传统支持向量机（SVM）方法中存在的不足，提出了基于数据优化法的SVM，它通过其它统计学模型优化训练数据集，进而提高分类器的辨识精度。

Considering the deficiency of traditional SVM , data optimization based SVM is presented . Its scheme is to improve the prediction accuracy of classifier by optimizing the training dataset with other statistical model .

我们先利用改进的K-means聚类方法从训练数据中获得基向量，再结合卷积神经网络提取字符图像的特征。

Firstly , we use the variant of K-means obtain the basis vectors from the training data , then extract character feature by combining convolutional neural network . 2 .

该方法综合了VQ和GMM的优点，通过用VQ误差尺度取代传统GMM的输出概率函数，减少了建模时对训练数据量的要求，提高了识别速度。

By adopting VQ error scale instead of probability output of tradition GMM , data capacity is reduced during modeling and high recognition rate is gained .

本文改进了现有的ICI和FCO自动迭代工具，用以收集机器学习的训练数据。

The thesis improves the existed ICI and FCO iterative compilation tool , and uses it to collect the machine learning samples .

对uCT图象进行图象增强、灰度校准以及三维图象配准后得到训练数据。

The training data is obtained by image enhancement , gray level calibration and three dimension registration .

再根据K-means方法将测试数据集分为同样多个聚类集，并通过欧式距离找到它们与训练数据子集之间的对应关系。

Then , test data set is divided into multiple cluster subsets by the K-means method . Find the corresponding relationship between the training data subsets and cluster subsets through Euclidian Distance .

此外，本文还重点介绍了基于GMM的说话人聚类算法，该算法聚类过程需要的训练数据少，聚类速度快。

In addition , this paper introduced the method of speaker clustering based on GMM , which required less speech data but faster speed during the process of clustering .

此时只需要很少的自适应训练数据就可以用基于Rosen梯度投影法的优化算法计算出线性组合中各码本的最佳权值。

An effective algorithm based on Rosen gradient projection method is developed to count the weight of each codebook in the linear combination .

改进的Bagging算法是正负图像中分别随机可重复的选取训练数据，这样就有效的避免了原始算法中的一边倒现象。

The improved Bagging algorithm can select training set respectively from positive and negative images randomly and repeatedly , which can avoid the " one-sided " phenomenon of original algorithm effectively .

本文通过RBF神经网络与免疫系统存在相似性，采用人工免疫原理对RBF神经网络训练数据集的泛化能力，动态调整神经网络隐层结构。

Through the similarity between radial basis function neural network and immune principle , use the generalization ability of artificial immune principle to RBF network data sets training , and adjust the structure of the neural network hidden layer .

自适应多阶Markov模型可根据训练数据与输入数据的相关度，自动适配合适阶数的Markov模型进行预测，从而在提高预测性能的同时降低训练数据质量对预测结果的影响。

The adaptive multi-order Markov model can improve prediction accuracy as well as reduce the influence of training data quality on prediction results by automatically adapting to appropriate order of Markov model according to the correlation between training data and input data .