Doesn’t matter who you are!
Sorts each item in the same way, no matter who you are
ex.No matter who goes to hacker news, the same thing is shown to the viewer
Takes who you are into account
Taking to account things I like and find other things that are similar
Can also use users that are similar to me to make recommendations
ex. Like Notebook & Titanic love story => Titanic is a good recommendation
Like Paranormal activity & Ring => Titanic is a bad recommendation
Matrix factorization(A.K.A SVD) and its variants like probabilistic matrix factorization to take recommendations to next level
How DL can be applied to recommendation system
-Easy to convert matrix factorization code into Keras script
-How we can modify our script to use DNN instead of just matrix factorization
-Borrow ideas from computer vision, especially residual networks and inception to see if we can improve upon a single neural network
-How unsupervised DL can be applied to build recommendation systems (auto-encoders & restricted Boltzmann machines)