Variational inference and variational self encoder


Variational inference and variational self encoder

Author: kailugaji blog Park

This paper mainly introduces the variable auto encoder (VAE) and its derivation process, but VAE involves some basic knowledge of probability and statistics, so in order to better understand VAE, it first introduces the variable inference and expectation maximization, EM) algorithm, and then introduce the variational self encoder, and give another understanding method (reference [3]).

1. Variational inference

2. Variational self encoder

Generally, only one sample can be taken, that is, k = 1. See reference [3].

3. Another understanding of variational autocoder: face to face joint distribution

4. References

[1] Changdecibels kailugaji blog Park

[2] Qiu Xipeng, neural network and deep learning [M]. 2019

[3] Article under VAE – scientific spaces

[4] Kingma D P , Welling M . Auto-Encoding Variational Bayes[J]. 2013.

[5] For details of variational inference, please refer to: Hua Junhao’s blog – variational inference, variational Bayes algorithm understanding and derivation