This is my exploration into the MovieLens data set. I use it to experiment with dimensionality reduction, embeddings, and using neural nets for collaborative filtering.
My approach to visualizing the data is detailed in this blog post
I begin by experimenting with different dimensionality reduction techniques, so that I can see how the movies are clustered, using PCA and then t-SNE.
My approach to the Neural Networks is detailed in this blog post
I begin by creating embeddings for the movies and for the users, and the use this as input to train an RNN. I also visualize the activatins of the RNN layers, to see which movies are most activating different nodes.
My use of word embeddings is detailed in this blog post
Finally, I use word embeddings to add an additional input to the neural network, in the form of word tags which users can add when they rate a movie.