Chapter 4: Relation and Matching Networks Using TensorFlow

1. A relation network consists of two important functions: the embedding function, denoted by 
   
   , and the relation function, denoted by 
   
   . 
2. Once we have the feature vectors of the support set, 
   
   , and query set,  
   
   , we combine them using an operator, 
   
   . Here, 
   
    can be any combination operator; we use concatenation as an operator to combine the feature vectors of the support set and the query set—that is, 
   
   .

3. The relation function, 
   
   , will generate a relation score ranging from 0 to 1, representing the similarity between samples in the support set, 
   
   , and samples in the query set, 
   
   .

4. Our loss function can be represented as follows:

    

   
5. In matching networks, we use two embedding functions, 
   
    and 
   
   , to learn the embeddings of the query set 
   
    and the support set 
   
   , respectively. 

6. The output, 

   

   , for the query point, 

   

   , can be predicted as follows: