目录
IV
目录
第 1 章 背景 ...................................................................................................................1
第 2 章 前置知识 ...........................................................................................................5
2.1 gamma 函数 ......................................................................................................5
2.2 二项分布(binomial distribution) ......................................................................6
2.3 beta 分布(beta distribution) ..............................................................................7
2.4 多项分布(multinomial distribution) ..............................................................10
2.5 狄利克雷分布(dirichlet distribution) .............................................................12
2.6 共轭先验分布(conjugacy prior) .....................................................................13
2.6.1 从二项分布到 beta 分布 ....................................................................14
2.6.2 从多项分布到 Dirichlet 分布 .............................................................16
2.7 总结 ................................................................................................................18
参考文献 ...............................................................................................................18
第 3 章 LDA 的 Gibbs Sampling 推导 ...........................................................................19
3.1 unigram 假设 ...................................................................................................19
3.2 Latent Dirichlet Allocation 介绍 ......................................................................21
3.3 马尔可夫链 Metropolis-Hasting Gibbs Sampling ................................25
3.3.1 马尔可夫链(markov chain) .................................................................25
3.3.2 Metropolis-Hasting 算法 ......................................................................27
3.3.3 Gibbs Sampling .....................................................................................29
3.4 伟大的采样公式： Collapsed Gibbs Sampling 采样公式推导 ....................30
3.5 总结 ................................................................................................................36
参考文献 ...............................................................................................................36
第 4 章 实现与应用 .....................................................................................................38
4.1 实现 ................................................................................................................38
4.2 应用 ................................................................................................................45
4.2.1 相似文档发现 .....................................................................................45
4.2.2 自动打标签 .........................................................................................47
4.2.3 LDA 与 LR（逻辑斯蒂回归）结合做新闻个性化推荐系统 ..............48
4.2.4 topic rank
[10]
..........................................................................................50
4.2.5 word rank ..............................................................................................52
4.2.6 文章质量评分算法 .............................................................................55
4.2.7 总结 .....................................................................................................59
参考文献 ...............................................................................................................59
第 5 章 并行化 .............................................................................................................61
5.1 AD-LDA .............................................................................................................61
5.2 spark-LDA .........................................................................................................63
5.2.1 切分块 .................................................................................................63
5.2.2 选择 .....................................................................................................65
5.2.3 计算和合并 .........................................................................................66
5.2.4 总结 .....................................................................................................67
参考文献 ...............................................................................................................67
第 6 章 变分贝叶斯的启蒙 .........................................................................................68
评论0