Description: 对五种典型的贝叶斯网分类器进行了分析与比较。在总结各种分类器的基础上，对它们进行了实验比较，讨论了各自的特点，提出了一种针对不同应用对象挑选贝叶斯网分类器的方法。 -Of five kinds of typical bayesian network classifiers are analyzed and compared. On the basis of summarizing the various classifiers, experiments have been carried out to compare them, discusses their features, put forward a kind of according to different application object selection method of bayesian network classifier.
Platform: |
Size: 6144 |
Author:楼梯 |
Hits:
Description: 这是matlab编写的3个常用机器学习分类器代码。其中包括了： 1）PCA 分类其；2）LDA分类器：3）naive贝叶斯分类器。 3个算法的实现参考了《Introduction to Machine
Learning》。 除了这3个分类算法的实现外，代码里面还包含了用于测试的main.m 主程序和一个实验的简要报告。实验在著名数据集acoustic_train_data 上进行。-This source code includes the implementation of three famous classifiers in machine learning. They are 1) PCA, 2) LDA and 3) Naive Bayesian. The detail theory behind these classifier can be found text book <<Introduction to Machine Learning>>. Besides the implementation of these 3 algorithms, the main.m for testing and a brief testing report is also included. (The testing is on data set acoustic_train_data) Enjoy:) Platform: |
Size: 371712 |
Author:hhj |
Hits:
Description: 感知行为的影响因素包括单个关节的动作和不同关节的组态。因此提出一种新的基于关节的位置差异的特征类型，联合包括静态姿势、动作、位移在内的行为信息进行识别。采用关节在两个时间和空间区域的差异来明确地模拟个别关节动力学和不同关节的组态。然后应用主成分分析（PCA）来获得所需的特征。同时应用非参数的简捷的贝叶斯最近邻（NBNN）分类器进行多类行为的分类。这个NBNN分类器避免了帧描述符的量化，计算“图像到类别”的距离而不是“图像到图像”的距离。15到20帧的数据就足以实现手势以及动作的识别，无需应用整个视频序列，极大地缩短了延时，提高了识别效率。-Perceived behavioral factors, including single-joint movements and configure different joints. Therefore suggests a new location-based differences in the characteristics of the type of joint, the joint including a static posture, movement, displacement including behavioral information for identification. Differences in the use of the joint region both time and space explicitly configured to simulate individual joints kinetics and different joints. Then principal component analysis (PCA) to obtain the desired characteristics. At the same time a simple nonparametric Bayesian nearest neighbor (NBNN) classifier to classify many types of behavior. This avoids the frame NBNN classification descriptors to quantify computing " image to the category" distance rather than distance " image to the image" of. Data 15-20 is sufficient to achieve the gesture and action recognition, without application of the entire video sequence, which greatly reduces the latency and improve the ef Platform: |
Size: 38912 |
Author:杨杨 |
Hits:
Description: 本程序是用作贝叶斯推论的，训练好分类器，用于计算联合概率密度实现推理-This procedure is used for Bayesian inference, well trained classifier is used to calculate the joint probability density to achieve reasoning Platform: |
Size: 1024 |
Author:zhouqian |
Hits:
Description: 模式识别的贝叶斯公式进行设计分类器A Bayesian formulation for large data pattern recognition is designed for classifier-A Bayesian formulation for large data pattern recognition is designed for classifier Platform: |
Size: 1024 |
Author:他里雾 |
Hits:
Description: 概率神经网络(Probabilistic Neural Network)是由D.F.Speeht博士在1989年首先提出，是径向基网络的一个分支，属于前馈网络的一种。它具有如下优点：学习过程简单、训练速度快；分类更准确，容错性好等。从本质上说，它属于一种有监督的网络分类器，基于贝叶斯最小风险准则。(The rate neural network, first proposed in 1989, is a branch of the RBF network and is one of the feedforward networks. It has the following advantages: the learning process is simple, the training speed is fast, the classification is more accurate, the fault tolerance is good, and so on. In essence, it belongs to a supervised network classifier based on Bayesian minimum risk criteria.) Platform: |
Size: 46080 |
Author:哼哼1214
|
Hits:
Search - Bayesian classifier