Computational methods for Bayesian clustering

贝叶斯聚类的计算方法

• 批准号: 2884420
• 金额: --
• 依托单位: University of Edinburgh
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2023
• 资助国家: 英国
• 起止时间: 2023 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2884420
• 关键词: Computational; methods; Bayesian; clustering

Bayesian clustering allows to partition data into groups of homogeneous points, while incorporating prior information on the data structure and on the partition itself. However, standard stochastic search methods for Bayesian inference of partitions are often computationally prohibitive. Optimization-based methods instead, while faster, often disregard posterior uncertainty. This project will investigate computationally efficient optimization methods for approximating the posterior distribution on partitions, while accounting for posterior uncertainty.

贝叶斯聚类允许将数据划分为同质点组，同时合并有关数据结构和分区本身的先验信息。然而，用于分区贝叶斯推理的标准随机搜索方法在计算上通常是令人望而却步的。相反，基于优化的方法虽然速度更快，但通常忽略后验不确定性。该项目将研究计算有效的优化方法，用于近似分区上的后验分布，同时考虑后验不确定性。