search

fengzhu1 / GA-DTCDR

This is the model in "A Graphical and Attentional Framework for Dual-Target Cross-Domain Recommendation" (IJCAI2020). GA-DTCDR is an optimized model for DTCDR ("DTCDR: A Framework for Dual-Target Cross-Domain Recommendation" in CIKM2019).
GNU General Public License v3.0
67 stars 18 forks source link

Common user #3

Closed Data-Designer closed 2 years ago

Data-Designer commented 2 years ago

Hi. I am confused about overlapping user IDs. In this code, I only see fully overlapping user IDs, not partial overlapping

Data-Designer commented 2 years ago

It seems that the source code does not distinguish between common user and distinct user, but directly integrates the representations of the two domains in Batch?

fengzhu1 commented 2 years ago

We have serialized the user IDs of the two domains. Common users from different domains have duplicate IDs, while for distinct users, their IDs are different. All users from different domains are in the same user-space. In this way, it will be easy to integrate the embeddings of common users. For example, if domain A has user ids: [1, 2], and domain B has user IDs [1, 3], then the joint user space is [1, 2, 3]. Thus, for domain A, we have three user embeddings [e_1^a, e_2^a, e_3^a], and for domain B, we have three user embeddings [e_1^b, e_2^b, e_3^b].

Data-Designer commented 2 years ago

Thanks. Your reply helps me a lot.