For AD-LDA/AD-HDP, the perplexity computation exactly follows that of LDA/HDP, because a single set of topic counts Nwk are saved when a sample is taken.
確定! 回上一頁