The Recommender Systems training takes 1 day.
The training starts with a general introduction, potential applications, and prerequisites for Recommender Systems. We then delve into the different types of models - Popularity-based baseline models, Content-based models, Collaborative Filtering models, and Hybrid models. We will explain each model in detail, compare them, and examine which one is best to use.
After that, we will do several lab exercises where we will apply two different types of Recommender Systems to the MovieLens dataset with movie ratings and compare how they perform. The goal of the training is for the student to be able to build a recommender system from scratch and evaluate what type of system is best to use in any particular situation.
The training includes theory, demos, and hands-on exercises. After this training you will have gained knowledge about:
- Applications of Recommender systems
- Required data and common considerations
- Types of Recommender systems
- Popularity model
- Content-based models
- Collaborative filtering
- Hybrid models
- Matrix factorization methods
- Dealing with changing contexts
- Exploitation vs exploration trade-off
- Performance evaluation: offline & A/B testing
- Evaluation metrics such as precision@k and recall@k
- Training and evaluating Recommender systems with LightFM