Using the same RBM setup mentioned in the preceding recipe, train the RBM on the user ratings dataset (trX) using 20 hidden nodes. To keep a track of the optimization, the MSE is calculated after every batch of 1,000 rows. The following image shows the declining trend of mean squared reconstruction errors computed for 500 batches (equal to epochs):

Looking into RBM recommendations: Let's now look into the recommendations generated by RBM-based collaborative filtering for a given user ID. Here, we will look into the top-rated genres and top-recommended genres of this user ID, along with the top 10 movie recommendations.

The following image illustrates a list of top-rated genres:

The following image illustrates a list of top-recommended genres: