Support Questions

Find answers, ask questions, and share your expertise
Announcements
Celebrating as our community reaches 100,000 members! Thank you!

Oryx: multi models

avatar
Explorer

Hello

 

I am using Oryx for collaboriative flltering on "user"-"item" association. I then call /pref/user/item and run /recommend/user to get recommended items for this users. I am using a couple of  Oryx serving/computation server (let s call it SERVER 1)

 

I want to add recommender for "user"-"user". I suppose I will run /pref/user/user2 and run /recommend/user/ to get recommended users for this user.

 

Therefore, do I have to setup a new couple of Oryx serving/computation server (SERVER 2)for this case OR i can reuse the SERVER 1 for this purpose ? If so, how can I distinguish item from user when doing /recommend/user ?

 

Regards

1 ACCEPTED SOLUTION

avatar
Master Collaborator

That is correct. The only "problem" is that the model has no idea what these entities are. It doesn't care. Right now your recommendations might include both users and items. That might be useful! but usually it's not. Something has to filter them. So, that also means you need a way to distinguish user and item IDs. 

 

You can filter them on the caller side. Just ignore values you don't like and go back and ask for more recs if there are not enough.

 

Or you can use RescorerProvider to inject the filtering logic onto the server side. That can be faster but is more work.

View solution in original post

1 REPLY 1

avatar
Master Collaborator

That is correct. The only "problem" is that the model has no idea what these entities are. It doesn't care. Right now your recommendations might include both users and items. That might be useful! but usually it's not. Something has to filter them. So, that also means you need a way to distinguish user and item IDs. 

 

You can filter them on the caller side. Just ignore values you don't like and go back and ask for more recs if there are not enough.

 

Or you can use RescorerProvider to inject the filtering logic onto the server side. That can be faster but is more work.