Bachelor
2025/2026
Recommender Systems
Type:
Elective course (Data Science and Business Analytics)
Delivered by:
Joint Department with MTS
Where:
Faculty of Computer Science
When:
4 year, 3 module
Open to:
students of one campus
Language:
English
ECTS credits:
4
Contact hours:
40
Course Syllabus
Abstract
The course trains specialists in designing and deploying modern recommender systems, combining a deep understanding of machine learning algorithms with solving real business problems. Participants will master methods for building personalized recommendations from classical approaches to neural network architectures and production-level deployment. The program emphasizes practical application: working with real-world data, optimizing infrastructure, and interpreting A/B test results.
Learning Objectives
- The course combines theoretical foundations with practical assignments involving modern libraries such as RecTools, LightFM, and PyTorch. This integrated approach enables students not only to master basic models but also to develop skills in using industrial tools. Upon completion, participants will be able to design and adapt recommendation algorithms for various business needs, interpret experimental results, and propose improvements for production systems.
Expected Learning Outcomes
- Explain the principles of recommender systems and their application in real products
- Gain skills in evaluating recommendation quality using accuracy metrics, ranking metrics, and beyond-accuracy approaches
- Implement basic models (KNN, PureSVD, ALS) and advanced neural architectures, including transformers.
- Set up validation pipelines and analyze metric trade-offs
- Use RecTools, LightFM, and PyTorch for building recommender systems
- Interpret A/B test results and select optimal solutions
- Propose improvements for production systems, including offline and online model updates and infrastructure considerations
Course Contents
- Introduction to Recommender Systems. Recommendation Quality Assessment Methods
- Heuristic-Based Models. Collaborative Filtering: KNN Approaches and Linear Models. Model Validation
- Collaborative Filtering: Matrix Factorization. Approximate Neighbor Search Lecture
- Hybrid Models Using Content Lecture
- Two-Stage Pipeline: Gradient Boosting Reranking Lecture
- Deep Learning in Recommender Systems
- Transformers in Recommendations
- Why Search Is 80% a Recommendation System
- Productivization of recommendations and infrastructure. Offline, nearline, and online recommendation scenarios. A/B testing
Bibliography
Recommended Core Bibliography
- Fault-tolerant systems, Koren, I., 2007
Recommended Additional Bibliography
- Parul Aggarwal, Vishal Tomar, & Aditya Kathuria. (2017). Comparing Content Based and Collaborative Filtering in Recommender Systems. Retrieved from http://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=edsbas&AN=edsbas.32D5064E