| date | duration | maintainer | order | title |
|---|---|---|---|---|
w10d2 |
60 |
todo |
10 |
Recommendation Systems |
- RecSys Lecture (30 min, check and personallize presenter notes for tips)
- Simple SVD Recommender lab (20 min)
- Surprise lab (20 min)
- (Optional) svdRec lab (15 min)
Remind the students often that RecSys is simply an application of Dimensionality Reduction methods to a business problem common to many industries. Students should have an intuitive understanding of latent features and quantification of objects into vector space.
- Introduce Content-Based and Collaborative Filtering methods
- Perform Collaborative Filtering from ratings matrices using
pandasandsklearnon movie data - Understand why this approach represents Collaborative Filtering, how Content-Based would differ, and how it might be implemented in a hybrid approach.
- Use the Surprise library that provides some nice built-in recommender functionality
- Understand how SVDs and other matrix decompositions are employed by recommender algorithms
- svdRec, a simple package created by one of our former instructors, Zach Miller
- Surprise, a full-featured approach
- 10 RecSys papers everyone should read