S*******t
发帖数: 97 | 1 最近在看Blei的Latent Dilechilet Allocation,看得一头
雾水。
1. Choose \theta = Dir(\alpha)
这个明白，\theta就是一个distribution,假设有K个topic,
\sum theta= 1;
2. for each of the N words w_n:
(a)choose a topic z_n = mult(\theta)
这个具体意思是什么？ 每个word对应一个topic
的概率？multinomial的话应该需要count啊，
从哪里来呢？比如P(topic1)=\theta_1, topic1
还需要一个count.
(b) choose a word w_n from p(w_n|z_n,\beta)
这个就更不清楚了，不是已经有word了吗？怎么又
choose一个word.
实在是搞不清楚。 | w********e
发帖数: 944 | 2 A topic is defined as a multinomial distribution over all words in the
vocabulary. In order to generate a word for the nth location in a document,
first select a topic (latent variable),then generate word according to the
multinomial distribution associated with the topic. |
|
