A Deep Clustering Network based on VaDE and GCN

• I improved the clustering accuracy by introducing GCN to VaDE. Dataset:USPS.
• The overall framework of the VGDCN. The design of the VGDCN refers to the architecture of SDCN. You can find more theoretical analysis in WDW_graduation_thesis.pdf.

• Run VGDCN_usps_GCN.py to get the experiment result.
• Run VaDE_usps.py to see the performance of VaDE on usps.

The clustering accuracy was improved by 5%. These are the images generated by VAE(Variational Autoencoder).

• This research is my undergraduate graduation project. Thanks for the support of my Alma Mater:Beijing Forestry University.
• I may continue to study the deep clustering in Shenyang Institute of Automation Chinese Academy of Sciences in my master's stage, which helped me a lot in this project. My new research will be published here in the name of Robot Vision Group.