NDBI021: User Preferences and Advanced Recommending Methods

Course Materials

Content of the course is available online: lecture slides, labs assignments, semestral work.

Course Language

The course will be taught in English as long as there is at least one non-Czech-speaking student.

Course Organization

The course will be divided into four main blocks: RecSys basics, user preferences, advanced recommending concepts, and domain-specific recommenders. The schedule is as follows:

Lectures

• Basics
  • 09.10 - Intro + basic RS
  • 16.10 - No content (RecSys conference)
  • 23.10 - Basic RS + open problems + highlights from RecSys
• User Preferences
  • 30.10 - User feedback variants (intro + explicit feedback)
  • 06.11 - User feedback variants (implicit + other feedback)
  • 13.11 - User preferences
• Advanced RS Concepts
  • 20.11 - Fairness, biases, and de-biasing
  • 27.11 - Multi-objective, multi-stakeholder RS
  • 04.12 - Dynamic RS, Reinforcement Learning, group RS - we only managed the Group RS part, Dynamic RS will be covered on 18.12.
  • 11.12 - Scalability, sparse linear autoencoders (guest lecture: Martin Spisak and Vojtech Vancura, Recombee)
• Domain-Specific RS
  • 18.12 - Intro + News domain (+ Dynamic RS) - we did not manage to get through News RS, these will be covered on 8.1. 2025
  • 08.01 - News + Peer recommenders + Music + E-commerce (Fashion) We'll see how much we can cover...

Labs

• Labs 1: investigate basic usage of a selected RecSys framework (assignment) + fill-in some preference data (v1, v2, v3).
• Labs 2: investigate advanced usage of a selected RecSys framework + report findings (assignment) + fill-in some preference data (different version than during previous labs; v1, v2, v3).
• Labs 3: investigate the stability and sequential biases in explicit feedback data (detailed assignment).
• Labs 4: investigate biases and fairness issues of your favorite recommending algorithms (detailed assignment, and starter pack).
• Labs 5: analyze existing system w.r.t. usage of personalization and recommendation; suggest improvements (assignment details).

Grading & Requirements

The following is required to obtain the credits:

• A research-oriented semestral work (TBD, probably focus on serendipity definitions).
• Participation in lectures and labs. If unable to attend regularly, please notify the instructor to discuss alternatives.

The exams will be oral with written preparation with rather broader questions to allow you to shape on what you remember - not otherwise.

Labs 1 - details

Part 1: Data Preparations for User Preference Processing

Download one of the movie lists available from here (NDBI021_lab1_ratings_v[1|2|3].docx). Fill-in your ratings and save for later

Part 2: Recommendation Frameworks

Select one of the frameworks listed in the file available here, explore the framework and complete the tasks suggested in the file. Prepare your notes for the next labs.

Deadlines:

• Finalize Part 1 by October 16 and Part 2 before the next labs (October 30).