15、遗传关联的潜在树模型森林研究

遗传关联的潜在树模型森林研究

1. 潜在树模型森林（FLTM）学习算法

FLTM学习算法的核心在于通过潜在变量捕获观测变量（如遗传标记）中的信息，并通过层次聚类逐步构建模型。以下是该算法的详细步骤：

Algorithm 1. FLTM model learning
INPUT: Xobs, a set of p observed variables (X = X1, ..., Xp),
D[X], the corresponding data for n subjects,
PartitionProc, a procedure to partition variables into non-overlapping clusters of variables,
τpairwise, a threshold used to guide PartitionProc,
Criter, a criterion to estimate information fading through the bottom-up model construction,
τlatent, a threshold used to constrain information fading,
α1, α2 and cardmax, parameters used to estimate the cardinality of latent variables.
OUTPUT: F and θ, respectively the graphical structure and the parameters of the FLTM model,
L, the whole set