Practical collapsed variational bayes inference for hierarchical dirichlet process
Practical collapsed variational bayes inference for hierarchical dirichlet process
复制标题
DOI：
10.1145/2339530.2339550
复制
发表时间：
2012-08
期刊：
影响因子：
--
通讯作者：
Issei Sato;Kenichi Kurihara;Hiroshi Nakagawa
中科院分区：
文献类型：
--
作者：
Issei Sato;Kenichi Kurihara;Hiroshi Nakagawa
针对层次Dirichlet过程（HDP）提出了一种新的塌陷变分贝叶斯（CVB）推理方法.虽然现有的CVB推理的HDP变量的潜在狄利克雷分配（LDA）是更复杂，更难实现比LDA，所提出的算法是简单的实现，不需要保持方差计数，不需要设置超参数，并具有良好的预测性能。
We propose a novel collapsed variational Bayes (CVB) inference for the hierarchical Dirichlet process (HDP). While the existing CVB inference for the HDP variant of latent Dirichlet allocation (LDA) is more complicated and harder to implement than that for LDA, the proposed algorithm is simple to implement, does not require variance counts to be maintained, does not need to set hyper-parameters, and has good predictive performance.