Hi, nice work about Variational Autoencoder on recommendation. However, I am confused about the method of data split.
In the preprocessing.py,
unique_uid = user_activity.index
unique_uid is the index of active user rather than the uid (unique_uid['userId']). Owing to the filter operator before, some userId are moved out. Then some valid userId at the end will not be considered if we adopt the index of user_activity rather than the actual uid. I guess it might be a error or is there any other meaning of that?
Looking forward to your reply, Thanks.
Best.
Hi, nice work about Variational Autoencoder on recommendation. However, I am confused about the method of data split.
In the preprocessing.py,
unique_uidis the index of active user rather than theuid(unique_uid['userId']). Owing to the filter operator before, someuserIdare moved out. Then some validuserIdat the end will not be considered if we adopt the index ofuser_activityrather than the actualuid. I guess it might be a error or is there any other meaning of that?Looking forward to your reply, Thanks.
Best.