Probabilistic Graphical Models Learning Records

2014/02/02

Probabilistic Graphical Models Assignment#1: IntroductiontoBayesianNetwork

• FactorProduct.m
• FactorMarginalization.m
• ObserveEvidence.m
• ComputeJointDistribution.m
• ComputeMarginal.m

涉及到Matlab的语法知识, 对课程本身的作业倒是比较容易. 这和第一次作业的目标是一致的:

The goal of this first assignment is for you to gain familarity with Bayesian networks and to understand how we might compute probability queries in these networks.

求联合概率分布时采用Factor Product的方法. 这把ComputeJointDistribution.m 和 FactorProduct.m 联系起来了.
求边缘概率分布时先求联合概率分布, 再做证据约减, 最后求边缘概率.

2014/02/03

ProbabilisticGraphicalModels Assignment#2: BayesNetsforGeneticInheritance

• phenotypeGivenGenotypeMendelianFactor.m

  根据孟德尔模型(mendelian model)计算每个factor的概率: P(phenotype | genotype)

• phenotypeGivenGenotypeFactor.m

  非严格的孟德尔模型(Non-MendelianModel).the alphaList stores the probabilities that a person with that genotype will have the physical trait

• genotypeGivenAlleleFreqsFactor.m

  根据等位基因组合产生的每个基因类型的概率

• genotypeGivenParentsGenotypesFactor

  已知父亲和母亲某个等位基因的分布, 求出下一代对应基因的概率分布.

• constructGeneticNetwork.m

• phenotypeGivenCopiesFactor.m

• constructDecoupledGeneticNetwork.m

• constructSigmoidPhenotypeFactor