Graph and Prototype-based

Cosine-Classifier and Contrastive Classifier

Contrastive Graph

Node and Edge co-embedding

Pareto DA, MetaAlign 

Knowledge Amalgamation 

• https://arxiv.org/pdf/2006.05525.pdf

• https://arxiv.org/pdf/2003.10477v4.pdf

•  

Chia data lam nhieu subset 

Dynamic Graph based for dynamic feature alignment 

 Graph Convolutional Networks (GCN) with vanilla Knowledge Distillation 

 training dung 2 Stage: stage 1 training GCN de tao ra pseudo labels , sau do stage 2 dung pseudo labels  train MLP (hierarchical) 

dung 2 optimization stages 1 time  hypothesis h for classification, the second hypothesis h* for domain adaptation (joint hypothesis)

training song song 2 model 1 model chi dung train pseudo label, 1 model train voi real labels 

MarkovGNN

• https://github.com/HipGraph/MarkovGNN

• 
  

Tri-Net   

• https://www.ijcai.org/proceedings/2018/0278.pdf

• https://books.google.co.kr/books?id=GS7RDAAAQBAJ&pg=PA224&lpg=PA224&dq=tri+classifier&source=bl&ots=PRvZF4FzhS&sig=ACfU3U3y2yPxiycHpD0R_b5Law5_z1ItTg&hl=en&sa=X&ved=2ahUKEwiKkbzk1536AhV1p1YBHeBdBNIQ6AF6BAgwEAM#v=onepage&q=tri%20classifier&f=false  

Bai toan ve Pareto va Evolutional computation 

Ensemble Graph

Video Domain Adaptation (AdaMAE)