search

lyst / lightfm

A Python implementation of LightFM, a hybrid recommendation algorithm.
Apache License 2.0
4.73k stars 691 forks source link

Item similarity without using `interactions` data? #535

Open econti opened 4 years ago

econti commented 4 years ago

Not an issue, just a question. In the fit method I see that user_features and item_features are optional.

fit(interactions, user_features=None, item_features=None, sample_weight=None, epochs=1, num_threads=1, verbose=False)

Wondering for a totally cold start problem where all I have are item_features if it is possible to fit a model, e.g. something like:

fit(item_features, interactions=None, user_features=None...)