New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
to_ratings_per_user function unclear #1
Comments
Hi @franziweindel, so basically the function constructs a matrix representing the mapping (u, i) -> v, where u is some user_id, i is the i-th interaction of the user with user_id u, and v is the item_id. The first dimension of the matrix is the number of users and the second dimension is the maximum number of interactions a user had. Since the matrix is rectangular, item_id -999 is used for padding. Does this explain it? |
Hi @FlorianWilhelm, Thank you for your response and explaination! It is clear now! However, in my end results I now get negative entries for the topic-item distribution phi, does that have something to do with the padding or the negative sampling for the BPR loss function. Thank you, Franzi |
Hi, it's the logits your are looking at. What are you trying to do btw? |
ahh okay thanks a lot! |
Nice! Let me know when your work is available, I am looking forward to read it. |
Hi Florian,
I am having troubles understanding what the function to_ratings_per_user does. Should it return the the number of total ratings each user has made or for each users the actual ratings?
Also when running it I do get an index error since both of the keys in the df dataframe that is first defined are floats, so they cannot be used to index into the cube variable.
Any help is appreciated, thanks,
Franzi
The text was updated successfully, but these errors were encountered: