I was reading through Recsim
Is there a simple implementation of this - On MovieLens data for example?
My understanding so far -
- The UserModel is for modelling and sampling users - based on user based features, latent and observable (?). This can be the demographics of the users and their behavioral data (like click/rating)
- Document Model - Features of the doc to be recommended. In this case it can be the movie genre, director etc.
- UserChoice Model - For generating a mapping user and document and simulating a response
Gaps in understanding -
- A simple implementation goes a long way in explaining. I am having trouble understanding the colab tutorials
- If I am generating a recommendation under certain constraints, how can that be incorporated? Example: If I am trying to push an email notification to a customer, I can't do that if (s)he is a DND customer (Do Not Disturb)
And the questions can go on, but all in all I think getting a wholesome simple explanation on this will go a long way.
P.S - For all recommender junkies, this is an extremely powerful tool. Please do check it out.