Recommendation algorithm from Twitter → tweet
Introduction
Introduction to the kind of datasets typical for collaborative filtering. Jeremy explains that we actually need some kind of vectors representing the tase of the user and the kind of the movie. With this information we could build a recommandation system. This leads us to latent factors.
Latent Factors & Excel
Step 1 of this approach is to randomly initialize some parameters. These parameters will be a set of latent factors for each user and movie. We can do it with Exel and in a Pytorch-Exel combination.
All the files can be found here. Here we get an intuitive understanding of latent space.
Embedding & Latent Factors
The Pytorch approach leads us to embeddings. We see that its actually a very simple idea.
(Later we build our embedding from scratch.)