(1)Jointly Optimizing Diversity and Relevance in Neural Response Generation
背景:
bland and generic responses.->diversify, cost of decreased relevance
以前使用CVAE实现diverse,VAE里面用高斯分布,高斯分布决定了diverse和relevant不可兼得
并且VAE存在一些问题
1)it suffers from the vanishing latent vari- able problem when applied to text generation tasks.->KL annealing method
2)VAE/CVAE conversation models can be limited to a simple latent representations such as standard Gaussian distribution->enrich the latent space by leveraging a Gaussian mixture prior
前人的方法(similiar to Luan2017):
(误)
现在的方法:
两个要点:
1.插值法,提出 Z i n t e r p Z_{interp} Zinterp = u Z S 2 S uZ_{S2S} uZS2S
latent 最近
最新推荐文章于 2023-03-19 15:19:54 发布
![](https://img-home.csdnimg.cn/images/20240711042549.png)