【PAPER-introduction】CausalVAE

Posted on Edited on In Views: 
CVPR 2021

CausalVAE:Structured Causal Disentanglement in Variational Autoencoder

Background knowledges

  • Disentangled Representation
    seperate the representation of features into individual fractors that are independent to each other.
    The other fractors keep unchanged and the one single fractor of the generated data changes while one fractor changes.
    It aims to find a lower dimentional, explainable abstact representation of high dimentional data.