Skip to content

yandexdataschool/recsys_course

Repository files navigation

YSDA Recommender Systems Course

This repository contains materials for the Recommender Systems course taught at the Yandex School of Data Analysis. This branch corresponds to the ongoing 2026 spring semester.

Syllabus

  • Week 1: Intro
    • Lecture: Course overview and organizational details, intro to Recommender Systems problem
    • Seminar: Basic recommenders, user-item latent space
  • Week 2: Candidate generation & metrics
    • Lecture: RecSys metrics & candidate generation: classic ML, ANN, mixing
    • Seminar: Yambda contest overview & baseline solution
  • Week 3: Ranking, diversity & metrics
    • Lecture: reranking - losses, algorithms, metrics; diversity control and MRR / DPP
    • Seminar: classic algorithms (MF, SLIM, EASE); ranking - pool building, undersampling, composite targets
  • Week 4: Deep learning for RecSys & neural candidate generation
    • Lecture: two-tower architecture, softmax model & sampled softmax loss, contrastive learning, negative sampling and LogQ
    • Seminar: paper review on LogQ correction and negative sampling techniques
  • Week 5: Neural candidate generation, pt.2
    • Lecture: cold start and long-tail, user / item encoding strategies, sequential models, beyond two-tower (GPU retrieval, generative retrieval)
    • Seminar: aspects of training neural networks for RecSys
  • Week 6: Neural ranking, pt.1
  • Week 7: Neural ranking, pt.2
  • Week 8: System Design, pt. 1
    • Lecture: Data architectures, logging, biases, data drift and monitoring
  • Week 9: System Design, pt. 2
    • Runtime design, candidate funnel, GPU inference, data delivery, controlled degradation, cold start
  • Week 10: RecSys Transformers applications
  • Week 11: Reinforcement Learning in RecSys
  • Week 12: Case Studies of Yandex's services
  • Week 13: Trends in RecSys

Staff

About

Recommender Systems course in YSDA.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors