question about some variables

[phamthanhtu310702]

I have just read the paper and moved into the source code. I was wondering what they are "psi", "H", "iw", and "vec", "star"
Could you kindly provide a explanation？
Thank you very much!! :)

[tdye24]

There are indeed some parts in the code that might be hard to understand, for instance, 'iw' stands for 'importance reweighting'.

I suggest delving into the following preliminary works and corresponding codes because I reused the markings from these codes while implementing SeeGera.

1. Title: Importance Weighted Autoencoders
   Paper link: https://arxiv.org/pdf/1509.00519.pdf

2. Title: Semi-Implicit Variational Inference
   Paper link: https://proceedings.mlr.press/v80/yin18b/yin18b.pdf
   Code link: https://github.com/mingzhang-yin/SIVI/blob/master/SIVAE.py

3. Title: Semi-Implicit Graph Variational Auto-Encoders
   Paper link: https://arxiv.org/pdf/1908.07078.pdf
   Code link: https://github.com/sigvae/SIGraphVAE/blob/master/sigvae.py

The latter two are particularly crucial. I believe that after familiarizing yourself with the preliminary works, everything will fall into place.

[akajinchen]

There are indeed some parts in the code that might be hard to understand, for instance, 'iw' stands for 'importance reweighting'.

I suggest delving into the following preliminary works and corresponding codes because I reused the markings from these codes while implementing SeeGera.

1. Title: Importance Weighted Autoencoders
   Paper link: https://arxiv.org/pdf/1509.00519.pdf
2. Title: Semi-Implicit Variational Inference
   Paper link: https://proceedings.mlr.press/v80/yin18b/yin18b.pdf
   Code link: https://github.com/mingzhang-yin/SIVI/blob/master/SIVAE.py
3. Title: Semi-Implicit Graph Variational Auto-Encoders
   Paper link: https://arxiv.org/pdf/1908.07078.pdf
   Code link: https://github.com/sigvae/SIGraphVAE/blob/master/sigvae.py

The latter two are particularly crucial. I believe that after familiarizing yourself with the preliminary works, everything will fall into place.

你好，代码中的k,j参数我没有太理解

[tdye24]

Thank you for your interest in our work.
K refers to the number of samples for importance re-weighting,
and J refers to the number of samples for variational distribution q.
Feel free to ask any questions.