Scholarlink

<p>Graphical modelling has become a useful tool in modern data mining. Graphical model selection by observations is an important challenge for various practical applications of this technique. Optimization-based approaches are a basis for many fast and efficient algorithms to solve this problem. Many known properties of graphical model selection algorithms are related to the case of the Gaussian distribution of the data. The case of non-Gaussian distributions is less investigated. An important issue is to understand how sensitive developed algorithms are to a change of distribution. In the present paper, we propose a simple methodology for comparative analysis of the sensitivity of graphical model selection algorithms to the change of distribution. This methodology is applied via extensive numerical experiments for the estimation of the dependency of the quality of chosen model selection algorithms on the change of distribution. Some interesting phenomena are observed and discussed.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

How Robust Are Different Versions of Graphical Model Selection Algorithms

  • 期刊论文

  • 发表日期：

    2025年12月5日

查看全文
  • Kostylev Ilya,
  • Kalyagin Valeriy

摘要

Graphical modelling has become a useful tool in modern data mining. Graphical model selection by observations is an important challenge for various practical applications of this technique. Optimization-based approaches are a basis for many fast and efficient algorithms to solve this problem. Many known properties of graphical model selection algorithms are related to the case of the Gaussian distribution of the data. The case of non-Gaussian distributions is less investigated. An important issue is to understand how sensitive developed algorithms are to a change of distribution. In the present paper, we propose a simple methodology for comparative analysis of the sensitivity of graphical model selection algorithms to the change of distribution. This methodology is applied via extensive numerical experiments for the estimation of the dependency of the quality of chosen model selection algorithms on the change of distribution. Some interesting phenomena are observed and discussed.