一种基于Dirichelt过程隐变量支撑向量机模型的目标识别方法
doi: 10.11999/JEIT140129
A Target Recognition Method Based on Dirichlet Process Latent Variable Support Vector Machine Model
-
摘要: 在目标识别中,对于样本数较多且分布复杂的数据,若将所有训练样本用来训练一个单一的分类器,会增加分类器的训练复杂度,且容易忽视样本的内在结构,不利于分类。因此人们提出了混合专家系统(ME),即将训练样本集划分为多个训练样本子集,并在每个子集上单独训练分类器。但是传统ME系统需要人为确定专家个数,并且每个子集的学习独立于后端的任务,如分类。该文提出一种基于Dirichlet过程(DP)混合隐变量(LV)支持向量机(SVM)模型(DPLVSVM)的目标识别算法,采用DP混合模型自动确定样本聚类个数,同时每个聚类中使用线性隐变量SVM(LVSVM)进行分类。不同于以往算法,DPLVSVM 将聚类过程和分类器的训练过程联合优化,保证了各个子集中样本的分布上的一致性和可分性,而且可以利用Gibbs采样技术对模型参数进行简便有效的估计。基于人工数据集、公共数据集以及雷达实测数据的实验验证了该文方法的有效性。
-
关键词:
- 目标识别 /
- 混合专家系统 /
- Dirichlet过程混合模型 /
- 隐变量支持向量机分类器
Abstract: In target recognition community, when dealing with large-scale and complex distributed data, it is very expensive to train a classifier using all input data and the underlying structure of the data is ignored. To overcome these limitations, the Mixture-of-Experts (ME) system is proposed, which partitions the input data into several clusters and learns a classifier for each cluster. However, in the traditional ME system, the number of experts are fixed in advance and clustering procedure and the classification tasks are de-coupled. To deal with these problems, a Dirichlet Process mixture of Latent Variable Support Vector Machine (DPLVSVM) is proposed. In DPLVSVM model, the number of clusters is chosen automatically by DP mixture model, and the linear Latent Variable SVMs (LVSVM) are employed in each cluster. Different from previous algorithms, in DPLVSVM, the clustering procedure and LVSVM are jointly learned to gain infinite discriminative clusters. And the parameters can be inferred simply and effectively via Gibbs sampling technique. Based on the experimental data obtained from the synthesized dataset, Benchmark datasets and measured radar echo data, the effectiveness of proposed method is validated.
计量
- 文章访问数: 2081
- HTML全文浏览量: 98
- PDF下载量: 775
- 被引次数: 0