基于潜狄利克雷分配模型的遥感影像聚类算法超参数敏感性分析
首发时间:2012-12-21
摘要:本文研究了基于潜狄利克雷分配(Latent Dirichlet Allocation,LDA)的遥感影像聚类算法中模型超参数的敏感性。首先,介绍基于LDA的遥感影像聚类算法中LDA模型的原理和超参数的性质,进而采用实验数据对模型的超参数进行实验与分析;然后,通过分析模型对超参数的敏感性和相关研究人员对超参数的估计提出的分析与假设,探讨超参数的估计方法,并通过超参数的估计来指导超参数敏感性分析实验的设计;最后,根据实验与分析,对模型超参数敏感性进行了综合分析与评估。
For information in English, please click here
The Sensitivity Analysis of Hyper-parameters in the Image Clustering Method base on Latent Dirichlet Allocation
Abstract:In this paper, the Sensitivity of Hyper-parameters in the image clustering method which base on the LDA model is analyzed. Firstly, the theory of image clustering methods based on LDA and the attributes of the Hyper-parameters are briefly introduced; then experiments about the Hyper-parameters are done using the image data and the effect is analyzed; thirdly, the estimation methods of the Hyper-parameters is discussed by using the experimental results and the related research; finally, the comprehensive evaluation of the sensitivity is done based on the experiments and analysis.
Keywords: Remote sensing image clustering LDA Hyper-parameters Sensitivity Analysiskey
论文图表:
引用
No.****
同行评议
共计0人参与
勘误表
基于潜狄利克雷分配模型的遥感影像聚类算法超参数敏感性分析
评论
全部评论0/1000