0
0

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?

More than 5 years have passed since last update.

思考: Disentangle Latent Representation of VAE

0
Posted at

刚刚读完Toward Controlled Generation of Text这篇paper,然后花了两个小时一直在思考: 这个模型跟GAN的区别到底在哪里?为什么GAN如日中天,VAE还经久不衰?难道只要加入一个中间embedding z,就能算是VAE么?最重要的是,从GAN的角度来思考,能不能对模型的本质达到更深的理解?

Screenshot from 2021-02-24 15-33-59.png

如上图所示,虽然比起2015年bowman的论文,确实算的上是一个不错的进展,但是,我们都知道图像到NLP领域就是降维打击——从论文引用数就能看出端倪,这样的成果实在不能说是惊人。

Screenshot from 2021-02-24 15-38-42.png

这是模型的结构。我们能看到潜在空间处多了一个c,然后还多了个Discriminators——注意这里的复数。

两个小时思考的结果是,一个简单的架构如下图——可行性未知,反正一定已经被人实验过了,只是我还没找到论文罢了。

Screenshot from 2021-02-24 15-47-53.png

难道不行吗????????? 我决定实验一下,冲冲冲。

0
0
0

Register as a new user and use Qiita more conveniently

  1. You get articles that match your needs
  2. You can efficiently read back useful information
  3. You can use dark theme
What you can do with signing up
0
0

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?