Question about items not appearing in train set

[miguelgfierro]

Hi @okuchaiev,

One question about the system. In the current implementation if an item does not appear in the training set, it is removed.

Is there other alternative to this? What would happen if we initialize the mappings with all train, validation and test set?

Thanks

[okuchaiev]

This is a cold-start problem. If the item or user wasn't seen before; the algorithm can not help at all. If we initialize mapping for training, valid and test with all items and all users, the items/user not seen in training examples, would just contain some random values (from initialization)

[miguelgfierro]

Thanks Oleksii