close
Paper Information
Li Wan et. al. “A Hybrid Neural Network-Latent Topic Model”, JMLR 2012
Introduction
Probabilistic graphical model 和 neural network model 都是機器學習裡很熱門的方法,本篇嘗試將hierarchical topic
model 結合 neural network. 借以整合各方長處: neral network提供低維的特徵表示給topic model以捕捉high-leval scene
structure.
Contribution
這篇paper提出neural network和topic model(LDA)結合的方法，使得整個架構可以用back-propagation去fine-tune。
Summarization
(a) Neural network (NN)
(b) Hierarchical topic model (HTM)
(c) Our hybrid model that combines the NN and HTM
training步驟分為三步
(a) Pre-training of neural network
(b) Pre-training of the hierarchical topic Model
(c) Joint optimization by gradient descent
Loss 如下：
Result
全站熱搜
留言列表
