Mathematical Sciences: Statistical Theory for Classification

首席调查员将为测试、评估和应用分类制定具体的概率模型。分类是构造相似对象的类的方法。为此目的建立的模型有混合模型、模式模型和进化分类的广义马尔可夫模型。研究人员介绍了一类新的概率模型--乘积划分模型，它在分类中有着广泛的应用，无论是在评估和选择分类方面，还是在使用分类进行预测方面都是如此。他还考虑了更专业的概率模型，用于设计集群存在的显著性检验，以及重建进化树。

The principal investigator will develop specific probability models for testing, evaluating and applying classifications. Classification is the method of constructing classes of similar objects. The established models for this purpose are mixture models, modal models, and for evolutionary classification, generalized Markov models. The investigator introduces a new class of probability models, product partition models, that have wide application in classification, both for evaluating and selecting classifications, and for using classifications in prediction. He also considers more specialized probability models for devising significance tests for the existence of clusters, and for reconstructing evolutionary trees.